Model training and experimentation
Training workflows, hyperparameters, checkpoints, experiment tracking, optimization, fine-tuning, and reproducibility for ML and LLM development.
40%
Best tweets about Machine Learning
50 Best Tweets About Machine Learning (2026)
Find the best tweets about machine learning, from models and datasets to training, evaluation, research papers, MLOps, and production systems.
Technical machine learning research, training, evaluation, data, engineering, deployment, and lessons from production.
- Creators
- 44
- Updated
Original Xholic analysis
What 50 top Machine Learning posts reveal
The feed emphasizes ML foundations, learning resources, and the practical lifecycle around data preparation, training, evaluation, deployment, and monitoring. Posts also surface a tension between increasingly autonomous ML workflows and the need for independent evaluation and deterministic verification.
- Dominant tone
- Positive
- Median score
- 33.7
- Leading format
- Announcement
- Recent posts
- 46%
60% of posts
All-time engagement
56% of posts
Published in 90 days
Conversation map
The themes creators return to
ML foundations and core algorithms
Mathematical and conceptual foundations of machine learning: probability, Bayes, information theory, KL divergence, optimization, gradient descent, calibration, and classical/deep-learning model families.
36%
ML learning roadmaps and educational resources
Books, courses, visual notebooks, skill trees, and step-by-step paths for learning machine learning, deep learning, and language modeling.
34%
Evaluation, verification, and reliability
Model and agent evaluation, task metrics, calibration limits, safety checks, deterministic verification, explainability, debugging, and production failure modes.
32%
LLM and generative AI engineering
Building, fine-tuning, evaluating, and serving LLM applications, including embeddings, RAG, tool calling, agents, tokenization, quantization, and inference optimization.
20%
ML data pipelines and quality
Data collection, cleaning, labeling, feature engineering, data contracts, feature stores, preprocessing, dataset quality, and drift.
16%
MLOps and production ML systems
Production lifecycle design for ML systems: CI/CD, training pipelines, model registries, monitoring, serving, continuous retraining, and operational reliability.
16%
ML systems and infrastructure
Compute, GPUs, distributed training, model-serving performance, hardware-aware development, latency prediction, and systems tooling.
12%
Tone and stance
Sentiment Positive leads
Author posture Supportive leads
Performance benchmark
- Median likes
- 60
- Median reposts
- 13
- Median replies
- 5
- Median views
- 6.9K
Posts with media make up 74% of this collection. Their median all-time score is 38.7, compared with 24.8 for text-only posts.
Format mix
- Announcement 56% · score 26.7
- List 20% · score 46.8
- Tutorial 18% · score 75.9
- Question 4% · score 4.64
Consensus and debate
Where creators agree, and where they do not
Shared view
Foundations remain a prominent learning topic
Bayesian reasoning, gradient descent, optimization, and first-principles visual materials are recurring topics in the feed’s technical learning content.
Shared view
ML is presented as a lifecycle
Several posts describe ML work from problem definition and data preparation through training, evaluation, deployment, monitoring, and retraining.
Shared view
Data quality is framed as upstream work
Production-oriented posts discuss data contracts, validation, curated data, feature stores, and drift monitoring, with data-quality controls placed before training and feature serving.
Shared view
Training operations support reproducibility
Posts discuss logging hyperparameters and periodic evaluations, resuming from checkpoints, testing training pipelines, model registries, and retraining triggers.
Open debate
Autonomous research versus independent verification
Posts about AI Scientist describe automated research workflows and peer-review-related milestones. Separate agent-evaluation commentary argues that systems should be evaluated as a model-plus-harness and checked with deterministic mechanisms outside the agent’s control.
Open debate
General LLM capability versus domain-specific design
A financial-ML post argues that off-the-shelf LLMs can confuse salience with materiality in financial tasks, while describing task-specific training and structured expectation modeling as approaches that improved results in the author’s prototypes.
Open debate
Complexity versus debuggability
A Google Search discussion notes that simpler linear systems can be easier to understand and debug than complex ML systems. Other posts describe agent systems that automate experimentation and parts of the ML lifecycle with human oversight at strategic points.
What performs
Patterns behind standout posts
Tutorials have the highest format-level median score
Tutorials have a median all-time score of 75.896, compared with 46.78 for lists and 26.72 for announcements. Tutorial evidence includes a Bayes explainer and visual ML notebooks.
Outliers span explainers, resource lists, and an announcement
The five identified score outliers are an AI-engineering term list, a book list, a Bayes explainer, an ML knowledge graph, and an automated-research announcement.
Posts with media show a higher median score
Media appears in 37 of 50 posts (74%). Posts with media have a median all-time score of 38.689, versus 24.774 for text-only posts.
Statistical standouts
- View standout post 1 Score 1458.3 · 43.31× median
- View standout post 2 Score 1240.3 · 36.84× median
- View standout post 3 Score 1144.1 · 33.98× median
- View standout post 4 Score 685.3 · 20.35× median
- View standout post 5 Score 468.1 · 13.9× median
Creator landscape
Who shapes this conversation
The five most represented creators account for 20% of the selected posts.
-
1. Vaishnavi
@_vmlops
2 posts
-
2. Aurimas Griciūnas
@Aurimas_Gr
2 posts
-
3. BURKOV
@burkov
2 posts
-
4. Shalini Goyal
@goyalshaliniuk
2 posts
-
5. Towards Data Science
@TDataScience
2 posts
-
6. Tivadar Danka
@TivadarDanka
2 posts
Aurimas Griciūnas links MLOps and data design
Aurimas Griciūnas’ two posts discuss ML CI/CD and training-pipeline artifacts alongside data contracts, feature stores, validation, and drift.
Vaishnavi covers fundamentals and ML systems
Vaishnavi’s posts highlight visual notebooks that derive ML algorithms and a curriculum covering systems infrastructure, hardware limits, and simulation.
Andriy Burkov’s posts emphasize code-first LLM learning
Posts about Andriy Burkov’s language-model book describe a concise, incremental path from earlier language-model approaches to Transformers, with PyTorch and Google Colab examples.
Since the previous snapshot
What changed since Aug 12, 2026
- 70% of the selected posts remained.
- The creator count changed by 0.
- The leading sentiment remained stable.
How this analysis was made
Themes, sentiment, stance, and post format are classified per tweet. All counts, shares, medians, creator concentration, freshness, and performance comparisons are then calculated directly from the published snapshot.
Xholic's all-time score compares engagement while accounting for reach, post age, and creator consistency. It is used for relative comparisons within this collection.
This report analyzes the exact 50-post snapshot shown below. AI identifies editorial categories and drafts explanations; all statistics are calculated from the snapshot, and every narrative claim is checked against cited posts before publication.
Best Machine Learning tweets
Top Machine Learning tweets from 44 creators
Ranked 01–50
- 01
@jahirsheikh8 ·
You’re not an AI Engineer until you understand these terms: • 🧠 Embeddings → Numerical meaning of text/data • 🔍 Vector DB → Similarity search storage • 📚 RAG → Retrieval-Augmented Generation • 🎯 Fine-Tuning → Task-specific model training • 🪶 LoRA → Lightweight fine-tuning method • ⚡ Quantization → Smaller/faster models • 🧵 Context Window → Model memory limit • 🔄 Function Calling → Structured tool usage • 🛡 Guardrails → Output constraints/safety • 📏 Eval Frameworks → Measure model quality • 🧮 Tokenization → How text becomes tokens • 🚀 KV Cache → Faster inference reuse • 🔥 Hallucination → Confident wrong output • 🪝 Prompt Chaining → Multi-step workflows Building demos is easy. Production AI is not.
- 02
@Zachly ·
You only need to read four books to truly get what’s going on in ML and data engineering: - Fundamentals of Data Engineering by Joe Reis - Designing Data Intensive Applications by Martin Kleppmann - AI engineering by Chip Huyen - Designing Machine Learning Systems by Chip Huyen If you read these four technical books and then read these four books on leadership and soft skills, you’ll be well on your way to massive success! - Radical Candor - Atomic Habits - How to Win Friends and Influence People - The Body Keeps Score What books would you recommend?
- 03
@techNmak ·
Most engineers have seen this formula. P(A|B) = P(B|A) × P(A) / P(B) Almost none can explain what it actually does. Here's Bayes' Theorem in plain English, and where it's hiding inside systems you use every day. The core idea in one sentence: Bayes' Theorem updates your belief about something after seeing new evidence. That's it. Four terms: Prior → what you believed before the evidence Likelihood → how probable the evidence is, given your hypothesis Evidence → how common the evidence is overall Posterior → your updated belief after seeing the evidence A concrete example: Say 40% of all emails are spam (your prior). You see a new email containing the word "lottery." 10% of spam emails contain "lottery." Only 1% of legitimate emails do. Plug into Bayes: P(spam | "lottery") = (0.10 × 0.40) / P("lottery") ≈ 87% The word "lottery" updated your belief from 40% → 87%. That's Bayes in action. Prior belief + new evidence = updated belief. Where it lives in AI: 1/ Spam filters The Naive Bayes classifier, the algorithm behind most spam filters - applies this exact calculation word by word across an entire email. Each word shifts the probability up or down. It's called "naive" because it assumes each word is independent of the others, which isn't realistic, but works remarkably well in practice. 2/ Medical diagnosis AI A patient has symptom X. What's the probability of disease Y? Bayes updates the base rate (how common the disease is) with the likelihood of seeing that symptom in patients who have it. Same formula, different domain. 3/ Your LLM's uncertainty Modern language models don't just predict the next token, they assign a probability to every possible token. The sampling process (temperature, top-p) is directly working with those probability distributions. Bayesian reasoning is embedded in every response your model generates. The insight most engineers miss: Bayes doesn't give you certainty. It gives you a rational way to update uncertainty. That's exactly why it's foundational to AI - real-world systems are never certain. They're always working with incomplete, noisy, probabilistic information. Every model that learns from data is, at its core, doing some version of this: Start with a belief. See evidence. Update the belief. That's Bayes. That's machine learning.
- 04
- 05
@SakanaAILabs ·
The AI Scientist: Towards Fully Automated AI Research, Now Published in Nature Nature: https://t.co/nNfpSV5e5I Blog: https://t.co/i6h8LVQOdl When we first introduced The AI Scientist, we shared an ambitious vision of an agent powered by foundation models capable of executing the entire machine learning research lifecycle. From inventing ideas and writing code to executing experiments and drafting the manuscript, the system demonstrated that end-to-end automation of the scientific process is possible. Soon after, we shared a historic update: the improved AI Scientist-v2 produced the first fully AI-generated paper to pass a rigorous human peer-review process. Today, we are happy to announce that “The AI Scientist: Towards Fully Automated AI Research,” our paper describing all of this work, along with fresh new insights, has been published in @Nature! This Nature publication consolidates these milestones and details the underlying foundation model orchestration. It also introduces our Automated Reviewer, which matches human review judgments and actually exceeds standard inter-human agreement. Crucially, by using this reviewer to grade papers generated by different foundation models, we discovered a clear scaling law of science. As the underlying foundation models improve, the quality of the generated scientific papers increases correspondingly. This implies that as compute costs decrease and model capabilities continue to exponentially increase, future versions of The AI Scientist will be substantially more capable. Building upon our previous open-source releases (https://t.co/H1tBT14Yx8), this open-access Nature publication comprehensively details our system's architecture, outlines several new scaling results, and discusses the promise and challenges of AI-generated science. This substantial milestone is the result of a close and fruitful collaboration between researchers at Sakana AI, the University of British Columbia (UBC) and the Vector Institute, and the University of Oxford. Congrats to the team! @_chris_lu_ @cong_ml @RobertTLange @_yutaroyamada @shengranhu @j_foerst @hardmaru @jeffclune
Watch video
- 06
@_vmlops ·
MACHINE LEARNING VISUALIZED: A FREE JUPYTER BOOK THAT SHOWS YOU HOW MODELS ACTUALLY LEARN open-source collection of notebooks by Gavin Hung, deriving ML algorithms from first-principles with gifs of them converging during training → gradient descent + optimizers → PCA + K-Means → perceptron + logistic regression → neural networks, backprop, loss landscapes → autoencoders + latent space math derivation + working code + visualization in every notebook. MIT licensed, open to PRs
Watch video
- 07
@sentient_agency ·
10 BOOKS SERIOUS AI RESEARCHERS ACTUALLY RECOMMEND (NOT THE ONES EVERYONE POSTS) Every AI reading list says the same five names. The people actually building these systems read deeper than that. Here's the shelf they point to when nobody's performing for an audience. 1. Probability Theory: The Logic of Science - E.T. Jaynes The book researchers quietly call life-changing. Jaynes reframes probability not as gambling odds but as the mathematics of reasoning under uncertainty, which is exactly what every modern model is doing. Dense, opinionated, and the closest thing the field has to a sacred text. Almost nobody outside the work has heard of it. 2. Information Theory, Inference, and Learning Algorithms - David MacKay The book that unites information theory and machine learning in one place, written by a Cambridge physicist who made it genuinely fun. Free online, full of puzzles, and on the shelf of nearly every researcher who actually understands why their models compress and predict the way they do. 3. Reinforcement Learning: An Introduction - Sutton and Barto The foundation under everything from AlphaGo to how modern models get fine-tuned with human feedback. Researchers don't recommend it because it's trendy. They recommend it because the ideas in it keep turning out to be the ideas that matter, decades later. Also free. 4. The Book of Why - Judea Pearl A Turing Award winner's argument that today's AI is stuck because it confuses correlation with causation, and a map of what real reasoning would require. The book that names the exact ceiling current systems keep hitting. Researchers cite it constantly. The public reads past it. 5. Vision - David Marr A neuroscientist's framework for how any system, brain or machine, processes information, written before deep learning existed and somehow predicting the questions it would face. The "levels of analysis" idea in here quietly shapes how serious people think about what a model is even doing. 6. Gödel, Escher, Bach - Douglas Hofstadter The cult book about how meaning and selfhood emerge from systems following simple rules. It won a Pulitzer and then got name-dropped to death, but almost nobody finishes it. The ones who do think differently about intelligence forever. The real one, not the summary. 7. Metaphors We Live By - Lakoff and Johnson The argument that human thought runs on metaphor, not cold logic, and that you can't build a mind on first-order logic alone. Researchers working on why language models grasp meaning the strange way they do keep circling back to this one. A genuine left-field pick. 8. The Society of Mind - Marvin Minsky One of AI's founding figures arguing that intelligence isn't one thing, it's a swarm of dumb little processes working together. Written as hundreds of one-page ideas. Out of fashion for years, now looking prophetic in the age of multi-agent systems. Insider catnip. 9. How to Solve It - George Pólya A 1945 book on mathematical problem-solving that quietly shaped how a generation of researchers think about breaking down hard problems, and that keeps surfacing in papers on how to make models reason. The bridge between human heuristics and machine reasoning. 10. The Mathematical Theory of Communication - Claude Shannon The original paper that invented information theory and, with it, the entire conceptual ground that machine learning stands on. Short, brutal, and foundational. Researchers revere Shannon the way physicists revere Newton. Most reading lists skip the source and quote the descendants. The popular books tell you what AI might do. These tell you how the people building it actually think. The difference is the whole point.
- 08
@TechWithKhushi ·
MACHINE LEARNING — MASTER TREE 🌲 Machine Learning │ ├── 01. Mathematics │ ├── Linear Algebra │ ├── Probability │ ├── Statistics │ ├── Calculus │ ├── Optimization │ └── Information Theory │ ├── 02. Python Foundations │ ├── NumPy │ ├── Pandas │ ├── Matplotlib │ ├── Seaborn │ ├── APIs │ └── Data Cleaning │ ├── 03. Data Preprocessing │ ├── Missing Values │ ├── Feature Engineering │ ├── Encoding │ ├── Scaling │ ├── Data Splitting │ └── Feature Selection │ ├── 04. Supervised Learning │ ├── Linear Regression │ ├── Logistic Regression │ ├── Decision Trees │ ├── Random Forest │ ├── XGBoost │ └── SVM │ ├── 05. Unsupervised Learning │ ├── K-Means │ ├── DBSCAN │ ├── Hierarchical Clustering │ ├── PCA │ ├── t-SNE │ └── Dimensionality Reduction │ ├── 06. Deep Learning │ ├── Neural Networks │ ├── CNNs │ ├── RNNs │ ├── LSTMs │ ├── Transformers │ └── Attention Mechanisms │ ├── 07. MLOps │ ├── Docker │ ├── Kubernetes │ ├── MLflow │ ├── Model Registry │ ├── Monitoring │ └── CI/CD for ML │ ├── 08. Generative AI │ ├── LLMs │ ├── Prompt Engineering │ ├── RAG │ ├── Fine-Tuning │ ├── Agents │ └── MCP │ ├── 09. Deployment │ ├── FastAPI │ ├── Flask │ ├── Cloud Deployment │ ├── APIs │ ├── Edge AI │ └── Inference Optimization │ └── 10. Future of ML ├── AI Agents ├── Multimodal AI ├── Robotics ├── Autonomous Systems └── AGI Research
- 09
@burkov ·
The Hundred-Page Language Models Book by Andriy Burkov is well regarded, and for a specific niche: readers who want to actually build a language model, not just read about one. Why it's good: - Density without fluff. True to the "hundred-page" branding, it moves fast through n-grams → RNNs → Transformers → LLM finetuning/prompting without padding. If you already know general ML but haven't sat down and coded a Transformer, this closes that gap efficiently. - Hands-on code. All examples run in PyTorch on Google Colab, so you're not just reading math — you build three different language model architectures yourself, including a Transformer from scratch. That's the book's real differentiator versus most "intro to LLMs" material. Endorsements from figures like Vint Cerf and Tomáš Mikolov (author of word2vec) call it clear and a solid starting point for language modeling, and reviewers like the CEOs of Weaviate, Qdrant, and LlamaIndex praise its concision and clarity for understanding how LLMs work under the hood. Burkov's prior book (The Hundred-Page Machine Learning Book) has a strong track record and is used as a university textbook, so there's real precedent for his teaching style working. Bottom line: if you want a fast, code-first path to genuinely understanding and building Transformer-based LLMs, and you're comfortable with Python and some math, it's a strong choice.
- 10
@SchmidhuberAI ·
Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!
Watch video
- 11
@Hesamation ·
KL divergence is a fundamental concept used in machine learning, from optimization a neural net to RL training of LLMs. but here is what it actually means:
Watch video
- 12
@Aurimas_Gr ·
As an 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿 you should also care about regular, non LLM-based ML models, productionising them comes with its own challenges. For example, 𝗖𝗜/𝗖𝗗 process is 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗳𝗼𝗿 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗣𝗿𝗼𝗷𝗲𝗰𝘁𝘀 compared to regular software. The important difference that the Machine Learning aspect of the projects brings to the CI/CD process is the treatment of the Machine Learning Training pipeline as a first class citizen of the software world. ➡️ CI/CD pipeline is a separate entity from Machine Learning Training pipeline. There are frameworks and tools that provide capabilities specific to Machine Learning pipelining needs (e.g. KubeFlow Pipelines, Sagemaker Pipelines etc.). ➡️ ML Training pipeline is an artifact produced by Machine Learning project and should be treated in the CI/CD pipelines as such. What does it mean? Let’s take a closer look: Regular CI/CD pipelines will usually be composed of at-least three main steps. These are: 𝗦𝘁𝗲𝗽 𝟭: Unit Tests - you test your code so that the functions and methods produce desired results for a set of predefined inputs. 𝗦𝘁𝗲𝗽 𝟮: Integration Tests - you test specific pieces of the code for ability to integrate with systems outside the boundaries of your code (e.g. databases) and between the pieces of the code itself. 𝗦𝘁𝗲𝗽 𝟯: Delivery - you deliver the produced artifact to a pre-prod or prod environment depending on which stage of GitFlow you are in. What does it look like when ML Training pipelines are involved? 𝗦𝘁𝗲𝗽 𝟭: Unit Tests - in mature MLOps setup the steps in ML Training pipeline should be contained in their own environments and Unit Testable separately as these are just pieces of code composed of methods and functions. 𝗦𝘁𝗲𝗽 𝟮: Integration Tests - you test if ML Training pipeline can successfully integrate with outside systems, this includes connecting to a Feature Store and extracting data from it, ability to hand over the ML Model artifact to the Model Registry, ability to log metadata to ML Metadata Store etc. This CI/CD step also includes testing the integration between each of the Machine Learning Training pipeline steps, e.g. does it succeed in passing validation data from training step to evaluation step. 𝗦𝘁𝗲𝗽 𝟯: Delivery - the pipeline is delivered to a pre-prod or prod environment depending on which stage of GitFlow you are in. If it is a production environment, the pipeline is ready to be used for Continuous Training. You can trigger the training or retraining of your ML Model ad-hoc, periodically or if the deployed model starts showing signs of Feature/Concept Drift. Let me know your thoughts. 👇
- 13
@heynavtoor ·
🚨 An AI just wrote a scientific paper. Came up with the hypothesis. Designed the experiments. Ran the code. Analyzed the data. Created the figures. Wrote every word. Then it passed peer review at a top machine learning conference. No human touched it. Not one word. Not one edit. This is not a demo. This actually happened. At ICLR 2025. It's called AI Scientist v2. An open source system that does the entire scientific research process. Autonomously. End to end. From idea to published paper. Here's what this system does on its own: → Generates research hypotheses from a broad topic you provide → Searches existing literature to check if the idea is novel → Designs experiments to test the hypothesis → Writes and debugs its own experiment code → Runs the experiments on GPUs → Analyzes the results with statistical methods → Creates publication-ready figures and visualizations → Writes the entire manuscript. Title to references. LaTeX formatted. → Reviews its own paper and improves it before submission Here's the wildest part: They submitted 3 fully AI-generated papers to an ICLR workshop. Reviewers were told some papers might be AI-generated but not which ones. One paper scored 6, 7, and 6 from three reviewers. That put it in the top 45% of all submissions. Above the average human paper. The AI outscored most human researchers. At a real conference. Through blind peer review. PhD programs cost $50,000 to $80,000 per year. Research takes 5 to 7 years. Postdocs earn $55,000 for more years of the same grind. 2.2K GitHub stars. Published research paper. Apache 2.0 License. 100% Open Source.
- 14
@FundamentEdge ·
Interesting paper from Bridgewater & Thinking Machines on judgment in financial tasks: natively, LLMs are quite poor here, but when trained, results can improve materially. Captures an issue inherent to machine learning & stocks: natively, machines are not good at identifying signal vs. noise & materiality. One of the first questions a good trader will ask is “what is the trigger that will move the stock?” Machines struggle intensely with this task, as the answer is often subtle or boring, and the question stocks ask changes dynamically with price & market regime. LLM’s have shown no real improvement vs. prior ML here: large language models aren’t sourcing materiality, they are sourcing salience, and treat lexical intensity as a proxy for significance, which is a really poor way to capture materiality. This is why out of the box earnings previews run through off the shelf LLMs are pretty bad / not useful at all for decision making. They require a lot of oversight, but in my prototypes, a few critical steps systematically breaking down key drivers that can either be earnings levers, narrative influences, or risk factors, set against a multi-dimensional expectations engine that include sell-side but also market inferred expectations improve results materially. Once you do this, you can tangibly see the vision elucidated in this paper on a company by company basis. But it’s obvious that an out of the box LLM trained on a web-scrape won’t (likely ever) have these capabilities. It is a scary idea to, as an investor, consider that perhaps with this approach I could put my portfolio on “auto-pilot” during earnings and the machine could make the right trades add/trim/enter/exit/flip with say a 60% batting average. Even a low-mid 50% earnings print move batting average that clips the tails of blow ups & short squeezes has the potential to so dramatically change the alpha generation approach at multi-managers. This is a problem I’ve been working on for almost 10 years in concept. It feels like it’s now becoming possible to productionize and scale. I unfortunately won’t share much of the specifics here (there are some aspects to this I consider proprietary), but this is an idea that we are working on co-designing with a few clients. Please reach out if it is interesting. https://t.co/lSt2vzPbXx
- 15
@Suhail ·
/goal for AI model training runs is *so* good - it really feels like the future. Very little babysitting now. Mine: Launch a full training run on 4 nodes. Continuously record things in an experiment document if it exists. Log hyper params, configs, periodic evals, performance insights, analyze training stability, and important changes for future analysis and reproducibility. Fix any major bugs you encounter while you monitor training but do not change the fundamental nature of the experiment without asking. If it crashes, resume again and keep training. Resume from latest reliable checkpoint you have. Reach <num> steps
- 16
- 17
@MIT_CSAIL ·
A free throwback MIT course breaking down how machine learning techniques can be applied to healthcare: https://t.co/TrQlckLh8o (v/@MITOCW) Here, MIT prof. & CSAIL principal investigator David Sontag discusses how AI can help sort thru medical data (Lecture 1).
Watch video
- 18
@andy_ai0 ·
1-month playbook to start learning AI In just one month, you’ll already be able to: - understand what AI, machine learning, deep learning actually mean - use beginner AI tools without feeling lost - understand core concepts like training data, models, overfitting - build a few small AI projects - decide which direction to go next Week 1: Build the foundation What to learn: - what AI is - AI vs machine learning vs deep learning vs generative AI - what a model is - what training means - supervised vs unsupervised learning - classification vs regression - training data, labels, and features - overfitting at a basic level - what neural networks are at a high level - what prompts, tokens, and hallucinations are What to do: - write short notes in your own words for the key concepts - take 3-5 real-life examples and identify inputs and outputs - test a few prompts in ChatGPT or Claude and observe where AI is useful and where it struggles Week 2: Start using AI hands-on What to learn: - Python basics - variables, lists, dictionaries, loops, functions - JSON basics - what APIs are - how AI tools receive input and return output - beginner prompt writing - how to give clear instructions - how to ask for structured answers What to do: - set up Python, Jupyter Notebook, or Google Colab - run simple Python exercises - make your first AI prompt experiments more structured - build tiny practice tasks such as: 1. text summarizer 2. bullet-point generator 3. study-note explainer Week 3: Learn the real ML basics underneath What to learn: - how machine learning models learn from data - loss and error at a simple level - train / validation / test split - accuracy, precision, and recall - why accuracy can be misleading - linear models and logistic regression at a beginner level - what embeddings are - semantic similarity - transfer learning - prompting vs retrieval vs fine-tuning What to do: - study one beginner dataset - understand what the input columns are and what the model predicts - compare simple examples of classification and prediction - make a short notebook or notes page explaining: 1. what data goes in 2. what the model tries to predict 3. how performance is measured 4. what can go wrong Week 4: Build small projects and choose your direction What to learn: - how to turn AI knowledge into simple projects - how to define input, process, output, and limitations - how to test outputs - how to improve prompts or workflows - the difference between learning AI for: 1. building apps 2. machine learning 3. deep learning 4. productivity/work use What to do: - choose 1-2 small projects and finish them - test them on real examples - write a short README or notes page for each project - reflect on which direction feels most interesting Everything here is best learned through practice For every concept you learn, try to: - explain it in simple words - test it with an example - build something tiny with it And from there, it becomes much easier to keep going without feeling overwhelmed
- 19
@mark_k ·
Recursive Self Improvement is here! Frontis-MA1 (35B) is a new AI4AI agent trained for recursive self-improvement in machine learning engineering. Full OpenMLE stack released. Model post-trained on four operators: Draft → Improve → Debug → Crossover. Learning and evolution close in one loop with real execution feedback. MLE-Bench Lite (12h/task, single 4090 @ 12GB): Base 39.4% → +Evo 60.6% → +Evo-Max 71.2% Beats GPT-5.5 + Codex. Gains transfer to held-out NatureBench. Weights + code fully open.
- 20
@goyalshaliniuk ·
Every AI System Is Built on Machine Learning Models From predicting trends to generating art, these 20 ML models are the real engines behind modern AI innovation. Here is how they shape the AI systems we use every day - 1. Regression & Classification Models • Linear & Logistic Regression - Predict numbers or classify binary outcomes for sales, spam, or fraud detection. • Decision Tree & Random Forest - Make hierarchical decisions and improve accuracy using ensemble learning. • Gradient Boosting (XGBoost/GBM) - Boost prediction performance by learning from past errors. • Support Vector Machine (SVM) - Separate data using hyperplanes for image or sentiment classification. • Naive Bayes - Classify text or emails using probability-based logic. 2. Clustering & Similarity Models • K-Nearest Neighbors (KNN) - Find similarities between data points for recommendations. • K-Means Clustering & Hierarchical Clustering - Group unlabeled data into clusters for segmentation or analysis. • Principal Component Analysis (PCA) - Reduce dimensions for faster visualization and better insights. 3. Neural Network-Based Models • Artificial Neural Networks (ANNs) - Mimic human brain patterns to detect complex relationships. • Convolutional Neural Networks (CNNs) - Analyze visual data for tasks like image recognition. • Recurrent Neural Networks (RNNs) & LSTMs - Understand sequential data like text, speech, and time-series. • Autoencoders - Compress and reconstruct data for anomaly detection and denoising. • Generative Adversarial Networks (GANs) - Create synthetic media and deepfake visuals. 4. Advanced Generative & Reinforcement Models • Transformers - Power LLMs like ChatGPT by understanding context in sequences. • Diffusion Models - Generate art, images, and videos from random noise. • Reinforcement Learning Models - Train agents to make decisions using rewards and feedback loops. Machine Learning Is the Core of Every AI Breakthrough Master these 20 models, and you will understand how AI learns, reasons, and creates the world of tomorrow.
Watch video
- 21
- 22
@alvarobartt ·
💥 Learn how to build your own tool-calling agent with @huggingface TRL + @Alibaba_Qwen Qwen3.5 on @Azure Machine Learning! - @NousResearch hermes-function-calling-v1, 500 single-turn samples - SFT with TRL on Qwen3.5 2B (released today!) on a single NVIDIA H100 - Everything on Azure, from Container Registry to Machine Learning! Step-by-step in the thread 🧵
- 23
@burkov ·
The Hundred-Page Language Model Book is Andriy Burkov's 2025 follow-up to his bestselling Hundred-Page Machine Learning Book, and it walks from ML basics through n-gram/count-based models, RNNs, Transformers coded from scratch in PyTorch, and finally LLMs with instruction finetuning and prompt engineering. Every code example runs on Google Colab, and the book progressively builds understanding from basic concepts to advanced architectures. Its central virtue is concision paired with a hands-on, code-first approach. Rather than front-loading theory and hoping you'll connect it to practice later, the book leans on PyTorch and Google Colab to give you an accessible path to actually implementing each idea. You don't just read about a Transformer—you build one, which is the difference between recognizing the architecture and understanding why each piece exists. The pedagogical structure is deliberately incremental. It introduces concepts gradually, starting with foundational count-based methods before advancing to neural and Transformer-based models, with each chapter building on the previous one. This matters more than it sounds: language modeling has a long conceptual lineage, and seeing the progression from n-grams to RNNs to attention gives you a mental map of why modern architectures look the way they do, instead of treating the Transformer as a magic black box that appeared from nowhere. That historical throughline is something many faster "just build an LLM" resources skip entirely. The concision itself is a feature, not just a gimmick. The hundred-page constraint forces ruthless prioritization—Burkov has a well-earned reputation for distilling large topics into their essential core, and the result is a book you can realistically finish in a weekend or a week of after-work reading. For a working developer or data scientist who wants a genuine mental model without committing to a 700-page textbook, that density-to-time ratio is the whole value proposition. It's also genuinely accessible on the math. The book presents mathematical foundations intuitively, using illustrations, diagrams, and worked examples rather than dense formalism, so you can follow the reasoning without a graduate background. It still respects your intelligence—it doesn't skip the math, it just makes it legible. Finally, the endorsements are heavyweight and not merely marketing filler. There's a foreword by Tomáš Mikolov, of word2vec and FastText fame, and back-cover praise from Vint Cerf, who called it a gem of clarity. Endorsements from the people who actually built the field carry more weight than the usual jacket blurbs, and they line up with the broader reception: readers working through it publicly have described it as a fantastic book. Net: it's a strong pick if you want a fast, buildable, intuition-first introduction to how LLMs work under the hood, and you have some Python plus a bit of math comfort. #LMtrainingData
- 24
@goyalshaliniuk ·
Ever wondered how machines actually learn from data? This step-by-step visual breaks down the 12-stage journey of Machine Learning, turning complex AI training into a simple, intuitive process anyone can follow. From defining the problem to collecting, cleaning, and labeling data… To training, tuning, and evaluating models… All the way to deploying it in the real world and retraining it over time... Each step plays a critical role in transforming raw data into real-world intelligence. Let’s quickly walk through it: 1. Define the Problem – What do you want the machine to solve? 2. Collect the Data – Pull in data from APIs, sensors, or databases. 3. Explore the Data – Understand structure, trends, and outliers. 4. Clean the Data – Fix missing, noisy, or incorrect entries. 5. Label the Data – Add correct answers if it’s supervised learning. 6. Split the Dataset – Into training and test sets (sometimes validation). 7. Choose the Algorithm – Like decision trees, regression, or neural networks. 8. Train the Model – Feed in the training data, let it learn patterns. 9. Evaluate the Model – Test on unseen data using metrics like accuracy. 10. Tune the Model – Improve it by adjusting parameters or trying variations. 11. Deploy the Model – Push it into production for real-world use. 12. Monitor & Retrain – Keep it updated and accurate over time. ✅ Key Takeaway: Machine learning is not magic — it’s methodical. And understanding this pipeline helps you build better models, make smarter decisions, and debug faster. 📌 Save this for your ML journey, whether you're just starting out or reviewing your fundamentals.
Watch video
- 25
@Aurimas_Gr ·
A breakdown of 𝗗𝗮𝘁𝗮 𝗣𝗶𝗽𝗲𝗹𝗶𝗻𝗲𝘀 𝗶𝗻 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗦𝘆𝘀𝘁𝗲𝗺𝘀 👇 And yes, it can also be used for LLM based systems! It is critical to ensure Data Quality and Integrity upstream of ML Training and Inference Pipelines, trying to do that in the downstream systems will cause unavoidable failure when working at scale. There is a ton of work to be done on the Data Lake or LakeHouse layer. 𝗦𝗲𝗲 𝘁𝗵𝗲 𝗲𝘅𝗮𝗺𝗽𝗹𝗲 𝗮𝗿𝗰𝗵𝗶𝘁𝗲𝗰𝘁𝘂𝗿𝗲 𝗯𝗲𝗹𝗼𝘄. 𝘌𝘹𝘢𝘮𝘱𝘭𝘦 𝘢𝘳𝘤𝘩𝘪𝘵𝘦𝘤𝘵𝘶𝘳𝘦 𝘧𝘰𝘳 𝘢 𝘱𝘳𝘰𝘥𝘶𝘤𝘵𝘪𝘰𝘯 𝘨𝘳𝘢𝘥𝘦 𝘦𝘯𝘥-𝘵𝘰-𝘦𝘯𝘥 𝘥𝘢𝘵𝘢 𝘧𝘭𝘰𝘸: 𝟭: Schema changes are implemented in version control, once approved - they are pushed to the Applications generating the Data, Databases holding the Data and a central Data Contract Registry. Applications push generated Data to Kafka Topics: 𝟮: Events emitted directly by the Application Services. 👉 This also includes IoT Fleets and Website Activity Tracking. 𝟮.𝟭: Raw Data Topics for CDC streams. 𝟯: A Flink Application(s) consumes Data from Raw Data streams and validates it against schemas in the Contract Registry. 𝟰: Data that does not meet the contract is pushed to Dead Letter Topic. 𝟱: Data that meets the contract is pushed to Validated Data Topic. 𝟲: Data from the Validated Data Topic is pushed to object storage for additional Validation. 𝟳: On a schedule Data in the Object Storage is validated against additional SLAs in Data Contracts and is pushed to the Data Warehouse to be Transformed and Modeled for Analytical purposes. 𝟴: Modeled and Curated data is pushed to the Feature Store System for further Feature Engineering. 𝟴.𝟭: Real Time Features are ingested into the Feature Store directly from Validated Data Topic (5). 👉 Ensuring Data Quality here is complicated since checks against SLAs is hard to perform. 𝟵: High Quality Data is used in Machine Learning Training Pipelines. 𝟭𝟬: The same Data is used for Feature Serving in Inference. Note: ML Systems are plagued by other Data related issues like Data and Concept Drifts. These are silent failures and while they can be monitored, we can’t include it in the Data Contract. Let me know your thoughts! 👇
- 26
@GithubProjects ·
Financial Machine Learning is a curated collection of resources and implementations for applying machine learning techniques to quantitative finance and investment strategies. - Integrates ML techniques with financial data for investment strategy development - Covers predictive modeling, satellite data analysis, and data imputation methods - Features collaborative research opportunities with quantitative hedge funds - Provides access to a daily research feed via https://t.co/hZVsRLgcuA
- 27
@shushant_l ·
I'm shocked most people use AI every day without knowing how AI actually learns. Here's the complete AI model training pipeline explained in one simple infographic. --- 📂 AI Model Training ┃ ┣ 📂 AI Training Basics ┃ ┣ 📂 Pattern Recognition ┃ ┣ 📂 Predictions ┃ ┣ 📂 Error Measurement ┃ ┣ 📂 Model Improvement ┃ ┗ 📂 Billions of Iterations ┃ ┣ 📂 Preparation ┃ ┣ 📂 Define The Goal ┃ ┣ 📂 Collect Data ┃ ┣ 📂 Clean The Data ┃ ┣ 📂 Remove Duplicates ┃ ┗ 📂 Filter Low Quality Data ┃ ┣ 📂 Data Processing ┃ ┣ 📂 Tokenization ┃ ┣ 📂 Convert Text To Tokens ┃ ┣ 📂 Numerical Representation ┃ ┣ 📂 Input Formatting ┃ ┗ 📂 Training Ready Data ┃ ┣ 📂 Model Architecture ┃ ┣ 📂 Choose Neural Network ┃ ┣ 📂 Transformer Architecture ┃ ┣ 📂 Initialize Model ┃ ┣ 📂 Random Weights ┃ ┗ 📂 Parameter Setup ┃ ┣ 📂 Pretraining ┃ ┣ 📂 Predict Missing Tokens ┃ ┣ 📂 Learn Language ┃ ┣ 📂 Learn Facts ┃ ┣ 📂 Learn Reasoning ┃ ┗ 📂 Discover Patterns ┃ ┣ 📂 Learning Cycle ┃ ┣ 📂 Calculate Loss ┃ ┣ 📂 Compare Predictions ┃ ┣ 📂 Backpropagation ┃ ┣ 📂 Update Weights ┃ ┗ 📂 Optimization ┃ ┣ 📂 Validation ┃ ┣ 📂 Test On Unseen Data ┃ ┣ 📂 Measure Accuracy ┃ ┣ 📂 Check Reasoning ┃ ┣ 📂 Evaluate Safety ┃ ┗ 📂 Test Generalization ┃ ┣ 📂 Post Training ┃ ┣ 📂 Supervised Fine Tuning ┃ ┣ 📂 Preference Optimization ┃ ┣ 📂 Safety Alignment ┃ ┣ 📂 Improve Helpfulness ┃ ┗ 📂 Improve Reliability ┃ ┣ 📂 Safety ┃ ┣ 📂 Reject Harmful Requests ┃ ┣ 📂 Protect Privacy ┃ ┣ 📂 Reduce Unsafe Outputs ┃ ┣ 📂 Follow Safety Policies ┃ ┗ 📂 Responsible Responses ┃ ┣ 📂 Evaluation ┃ ┣ 📂 Math ┃ ┣ 📂 Coding ┃ ┣ 📂 Science ┃ ┣ 📂 Language ┃ ┣ 📂 Logic ┃ ┗ 📂 Real World Tasks ┃ ┣ 📂 Deployment ┃ ┣ 📂 Optimize Inference ┃ ┣ 📂 Efficient Serving ┃ ┣ 📂 Production Release ┃ ┣ 📂 User Access ┃ ┗ 📂 Scalable Infrastructure ┃ ┣ 📂 Continuous Improvement ┃ ┣ 📂 Collect Feedback ┃ ┣ 📂 Fix Weaknesses ┃ ┣ 📂 Improve Data ┃ ┣ 📂 Enhance Safety ┃ ┗ 📂 Release Updated Models ┃ ┣ 📂 End To End Pipeline ┃ ┣ 📂 Goal ┃ ┣ 📂 Data Collection ┃ ┣ 📂 Data Cleaning ┃ ┣ 📂 Tokenization ┃ ┣ 📂 Model Initialization ┃ ┣ 📂 Pretraining ┃ ┣ 📂 Loss Calculation ┃ ┣ 📂 Backpropagation ┃ ┣ 📂 Optimization ┃ ┣ 📂 Validation ┃ ┣ 📂 Post Training ┃ ┣ 📂 Safety Alignment ┃ ┣ 📂 Evaluation ┃ ┣ 📂 Deployment ┃ ┗ 📂 Continuous Improvement ┃ ┗ 📂 Key Takeaways ┣ 📂 AI Learns Patterns ┣ 📂 Data Quality Matters ┣ 📂 Pretraining Builds Knowledge ┣ 📂 Post Training Improves Safety ┣ 📂 Massive Compute Is Required ┗ 📂 Learning Never Truly Stops
- 28
@toly ·
MetaTimer: Using Large Language Models for Precise, Prompt-Aware Inference Latency Prediction The rapid proliferation of large language models (LLMs) in production systems has exposed a fundamental limitation: inference latency varies dramatically across prompts due to differences in semantic complexity, required reasoning depth, output length, and generation dynamics. Conventional prediction methods—ranging from token-count heuristics and hardware Roofline models to traditional machine-learning regressors—fail to generalize because they cannot capture these prompt-specific nuances. Accurate a priori estimation of processing time is essential for resource scheduling, dynamic batching, cost forecasting, service-level guarantees, and user-experience enhancements. We introduce MetaTimer, the first framework to repurpose a lightweight LLM itself as a high-precision meta-predictor capable of forecasting the exact wall-clock inference duration required by any target LLM for an arbitrary input prompt. A compact 8B-parameter model is fine-tuned on a massive corpus of millions of prompt–execution pairs collected across heterogeneous model families (GPT-4-class, Llama 3.1, Claude, Mistral), quantization levels, decoding strategies, and hardware accelerators. The predictor employs chain-of-thought reasoning to decompose prompt semantics, estimate output token distributions and reasoning trajectories, and integrate model- and hardware-specific performance profiles, yielding fine-grained predictions for Time-to-First-Token (TTFT), Time-Per-Output-Token (TPOT), and total latency. Extensive evaluations on held-out benchmarks spanning reasoning, creative writing, coding, and long-context tasks demonstrate state-of-the-art accuracy: a mean absolute percentage error (MAPE) of 6.3% for end-to-end latency—representing a >40% reduction in mean squared error relative to the strongest Roofline–ML baselines—and strong zero-shot generalization to unseen models and platforms. When integrated into production serving stacks (vLLM, TensorRT-LLM, Triton), MetaTimer delivers up to 31% gains in resource utilization and tail-latency reduction. These results establish that LLMs possess emergent capabilities for computational self-modeling, opening a new paradigm for self-aware, adaptive, and energy-efficient generative AI infrastructure. We publicly release the predictor model, dataset, and serving plugins to accelerate research in meta-performance modeling for frontier AI systems.
- 29
@eric_seufert ·
I'm happy to share some of the research that I've been working on today: As ad platforms become more opaque and automated end-to-end, advertisers are left with few levers of control over campaign performance. I wanted to interrogate an idea: could advertisers treat "black box" platforms as teacher models and use a behavioral distillation process, focusing on multimodal creative and ad context, to predict ROAS performance for ad instances? And if so, what kind of model could best express the interactions between those features? Most machine learning research focused on digital advertising is published by the largest platforms themselves. I wanted to address this asymmetry and explore what advertisers could build independently to improve performance within this increasingly automated environment. That idea ultimately became DeCANT: a Deep Creative Attention-based Network for pre-Testing. The model architecture uses self- and cross-attention to condition the semantic interpretation of a creative on the environment in which it is deployed. The model architecture supports multimodal ad creative and predicts ROAS on a context-conditioned basis. Operationally, a model like DeCANT can fit into a pre-testing regime as an automated filter: creatives are produced through a generative pipeline, and the student model that learns the process for the given advertising channel is invoked on the proposed creative and context. The model produces an expected ROAS, which is compared against the advertiser's testing threshold to determine whether the creative is uploaded to the platform. https://t.co/TgElrkFWwI
- 30
@Suryanshti777 ·
Holy shit… someone just made machine learning click. Not static diagrams. Not math-heavy PDFs. Not black-box training. Real algorithms — training step-by-step — visually. It’s called Machine Learning Visualized and it lets you watch models learn in real time. Here’s why this is different: Instead of dumping theory first, it shows optimization happening live: • gradients moving • weights updating • decision boundaries shifting • loss decreasing • models converging You literally see learning happen. Everything is built from first principles: • Gradient Descent • Logistic Regression • Perceptron • PCA • K-Means • Neural Networks • Backpropagation No magic. Just math → code → visualization. Each chapter is a Jupyter notebook that derives the math then implements it then animates training. So you can watch: • neural nets shape decision surfaces • PCA rotate feature space • K-means clusters form live • gradient descent find minima • sigmoid reshape boundaries • backprop update weights step-by-step This solves a huge problem: Most ML resources teach: math → code → ??? → trained model This shows: math → code → learning process → result Which means you finally understand: • why gradients matter • how weights evolve • what loss landscapes look like • how convergence actually happens • why deep nets learn non-linear functions Even better: You can open any notebook modify parameters and watch behavior change instantly. Learning ML becomes interactive. Not passive. Not abstract. Not confusing. Just… visible. Perfect for: • beginners learning ML • devs moving into AI • interview prep • teaching concepts • understanding backprop • visual learners • building intuition This is the kind of resource that makes neural networks finally “click”. Link: https://t.co/5wWTKZPJHE We’re moving from: reading about ML → watching ML learn That’s a big shift. Because once you can see training, you stop memorizing… and start understanding. AI education just got visual.
Watch video
- 31
@techyoutbe ·
This is how ML models actually go live 🚀 (Stop guessing, follow this ML flow) Raw Data Collection → Gather data from multiple sources Data Cleaning → Remove errors, duplicates, missing values Data Transformation → Convert data into usable format Feature Engineering → Create meaningful input features Model Selection → Choose best algorithm for problem Model Training → Train model on prepared dataset Model Evaluation → Check accuracy and performance metrics Model Deployment → Push model to production environment Model Monitoring → Track performance and detect issues Model Tuning → Improve model with adjustments Result → Process complete with stable model
- 32
- 33
@agenticgirl ·
Every AI engineer eventually builds a small personal library. Not tutorials. Not random blog posts. But the books that actually explain how modern AI systems work. Here are some of the titles you’ll find in many AI engineers’ bookshelves: ➜ Hands-On Machine Learning : Aurélien Géron A practical bridge between ML concepts and real implementations. https://t.co/ejDrBquAIz ➜ Build a Large Language Model (From Scratch) : Sebastian Raschka A clear breakdown of how LLM components actually work. https://t.co/ElfjNLejYt ➜ Deep Learning : Goodfellow, Bengio & Courville One of the most respected references in modern deep learning. https://t.co/4G8AkBgajO ➜ AI Engineering : Chip Huyen How real AI applications are built using foundation models. https://t.co/YVYyIJScmo ➜ Designing Machine Learning Systems : Chip Huyen Covers monitoring, data drift, and production ML challenges. https://t.co/uUn9TgNGuN ➜ Mathematics for Machine Learning : Deisenroth, Faisal & Ong The math foundations behind ML algorithms. https://t.co/6CI9eAhhzv ➜ Hands-On Large Language Models : Jay Alammar & Maarten Grootendorst Great visual explanations of transformers and LLM workflows. https://t.co/NAV0UAjNaO ➜ Practical MLOps : Noah Gift & Alfredo Deza How to deploy, scale, and maintain ML systems. https://t.co/FLtCt4uWyA ➜ Prompt Engineering for LLMs : John Berryman & Albert Ziegler Structured prompting techniques for real LLM applications. https://t.co/A0h8X0uOQz If you're serious about AI, start building your own AI engineering library.
- 34
@_vmlops ·
THIS GITHUB REPO IS A TEXTBOOK, A FRAMEWORK, A SIMULATOR, AND A HARDWARE LAB - ALL ONE CURRICULUM harvard-edge built the whole AI engineering stack in one repo, not five disconnected projects → vol I + vol II: single-machine systems through distributed, at-scale infra → tinytorch: build your own ML framework across 20 modules → hardware kits: deploy to arduino, raspberry pi, real memory/power limits → mlsys·im: simulate infrastructure most people can't afford to rent → staffml: physics-grounded interview prep for ML systems roles
- 35
@Al_Grigor ·
LLM systems feel like a new paradigm. In practice, much of the lifecycle still follows patterns that existed long before generative AI. One useful lens is CRISP-DM, a framework originally designed for data mining projects and widely adopted in data science. Even though the tools have changed, its phases map surprisingly well to how modern AI systems are built. Here is how the typical stages compare. 1. Business Understanding - Traditional ML: define the prediction task and success metrics. - AI systems: define the AI-powered product use case and the user experience you want to enable. 2. Data Understanding - Traditional ML: explore labeled datasets, distributions, and features. - AI systems: identify the inputs your system will use such as documents, images, APIs, databases, or external tools. 3. Data Preparation - Traditional ML: feature engineering, cleaning, and dataset curation. - AI systems: chunking documents, generating embeddings, building indexes, and wiring tools for agents. 4. Modeling - Traditional ML: train and tune models on structured datasets. - AI systems: prompt design, schema definition, retrieval pipelines, and agent behavior. 5. Evaluation - Traditional ML: metrics like accuracy, precision, and recall. - AI systems: task success, human feedback, and observable system behavior. 6. Deployment - Traditional ML: model serving and batch or online inference pipelines. - AI systems: full AI-powered applications that combine models, tools, and orchestration. The techniques look different, but the lifecycle remains largely the same. This is one reason many data scientists can smoothly transition into AI engineering roles. Read more about how CRISP-DM applies to AI Engineering: https://t.co/rkQw4fzVmT
- 36
@panditdhamdhere ·
If you're building AI applications in Rust, these are five of the strongest libraries to learn by the end of 2026. 🦀 → Burn - Deep learning - native Rust framework with training, inference, GPU acceleration (CUDA, WGPU), autodiff, modular design. Great alternative to PyTorch for Rust developers. → Candle - LLM - inference - Lightweight framework from Hugging Face. Excellent for running Llama, Mistral, Phi, Qwen, Gemma, and other transformer models efficiently in Rust. → tch-rs- PyTorch bindings- Rust bindings for LibTorch. Ideal if you're migrating PyTorch code to Rust or want access to the PyTorch ecosystem. → Ort (ONNX Runtime) - Production inference- High-performance ONNX Runtime bindings. Deploy models exported from TensorFlow, PyTorch, or scikit-learn with excellent speed. SmartCore - Classical machine learning- Pure Rust machine learning library supporting regression, classification, clustering, PCA, KNN, Random Forests, SVMs, and more. Great for non-deep-learning tasks. If your goal is AI agent development. A modern Rust AI stack. LLM Framework → Rig Local Model → Candle Tokenization → tokenizers Embeddings → FastEmbed Vector Database → Qdrant or LanceDB API → Axum Async Runtime → Tokio
- 37
@ForwardFuture ·
“I rewrote my code in 30 minutes — and it ran 200× faster.” @ctnzr VP, Applied Deep Learning Research @NVIDIA on the moment GPUs changed everything: “NVIDIA showed up in our lab and said, ‘You should try CUDA.’ I plugged in a GPU, rewrote my SVM training code, and it ran 200× faster than my CPU version.” “I thought, that’s it. This is dramatically easier and clearly the future of machine learning.” “The vision was simple: accelerate the world’s most important computations by 10× or 100×, and use that to power AI.” “The compute required for intelligence is essentially unbounded.”
Watch video
- 38
- 39
@GaryMarcus ·
This was right five years ago, and still is: “Large scale pretrained models are certainly likely to figure prominently in artificial intelligence for the near future, and play an important role in commercial AI for some time to come. The results that have been achieved with them are certainly intriguing and it is worthwhile pursuing them. But it is unwise to assume that these techniques will suffice for AI in general. It may be an effective short-term research strategy to focus on the immediate challenges that seem to be surmountable, but focusing on the surmountable may not get us to what is most necessary: a firm foundation for reliably integrating statistics and machine learning with reasoning, knowledge, common sense and human values.” Gary Marcus and Ernest Davis, 2021, “Has AI found a new Foundation?” at https://t.co/9VqBTg0FtK
- 40
@WevolverApp ·
Engineering, manufacturing, and architectural design data, by contrast, are fragmented, fiercely proprietary, and rarely centralized, even within a single department of an organization. For machine learning to deliver value in these domains, significant groundwork must be laid: cleaning, labeling, and structuring datasets that were never intended for algorithmic use. This includes everything from simulation outputs and CAD files to process metadata and real-world performance feedback. Learn more: https://t.co/giaPRtfJkU #engineering #technology
Watch video
- 41
@heyrimsha ·
This is genuinely cool. A 17-year-old high school student built an AI system that analyzes retinal images for patterns linked to autism and ADHD. His name is Edward Kang. The project is called RetinaMind. It won him $175,000 at the 2026 Regeneron Science Talent Search. Here is why the whole thing works at all. The retina, the layer of light-sensing tissue at the back of your eye, is not really a separate organ. In an embryo it grows out of the same tissue as the brain. Neurologically speaking, the retina is a piece of the brain that got pushed forward to catch light. Anything that changes the way the brain wires itself up early in development is likely to leave a fingerprint in the retina too. Researchers had already suspected this. A team at the Chinese University of Hong Kong published a paper a few years ago showing you could train a model on retinal photos and separate autistic children from neurotypical children at meaningful accuracy. Kang read that paper as a school project three years ago. He said later he thought it was fascinating and unintuitive that you could look at the eye and learn something about the brain, and he set out to improve on it. He taught himself the coding and machine learning he did not already know. He pulled a large public dataset of retinal images. He trained a suite of models, combined the best ones, and got RetinaMind up to around 89 percent screening accuracy for autism and ADHD from a single retinal image. That is the part that would have been enough to win a science fair. He did not stop there. He took a technique called GradCAM, an explainability method that shows you which pixels the model actually looked at to make its prediction, and ran his own model through it. That gave him a heat map, a red overlay on the retina showing exactly which regions of the eye were driving the diagnosis. Then he did the thing almost nobody at his age does. He asked why. If the model is looking at those regions, what is actually different about the cells there? He built a retinal cell model in a lab. He started studying gene changes in the regions the AI kept highlighting, looking for the biological reason those areas might develop differently in kids with autism or ADHD. Then he validated what he found in a second cell model. That is the arc that impressed the judges. He built the AI. He asked the AI what it was looking at. Then he went into a wet lab to check whether the biology backed it up. The caveats matter. Independent experts have pointed out that the retinal differences RetinaMind picks up on may not be specific to autism or ADHD, and could reflect a broader category of neurodevelopmental variation. Kang has said the same thing, and is now working on a version of the model that could distinguish severity levels. This is not a diagnostic replacement. It is a screening tool, and it is early. But the mindset behind the whole project is the part worth stealing. Start with a real question you actually want the answer to. Use AI to see patterns in data that a human eye would not catch. Then and this is the step almost everyone skips, ask the model what it is looking at, and go check whether the underlying reality really works that way. Most of the interesting AI work of the next ten years is going to look like this. Not "an AI that does X." An AI that finds something no human noticed, plus a human who followed the trail back to why it is there. He is 17. He has three years of this behind him already.
- 42
@glenngabe ·
Some interesting nuggets from the latest Search Off The Record Podcast about Google using AI in Search rankings. Here is Google's Nikola Todorovic, Director of Software Engineering at Google Search, about how AI is used to impact rankings and why simpler linear systems are sometimes easier to debug than more complex AI systems: Nikola: "The reason it's not so easy to apply AI everywhere (in Search) is because the models function like a black box. You don't always understand what's happening underneath. It's a complex set of neural networks. The linear models are the easiest ones to understand and debug, because it's not like you can just put your AI or ML system into search and reap the most benefit from your side by side experiments." "Then you will get to something and launch it, but you will have problems with that as well because maybe the systems evolved, the searches evolve, and so on. And then you will need to debug this and replace it (at some point). And this kind of replacement and changes is complicated. So the more you can understand how these things work, which signals you are using, which signals are important for relevance, for quality, for the safety of the results, (the easier it is to debug). So you do need to understand the system and the more complex the AI or the ML systems, the more challenging it is." https://t.co/Pf5stUq4HZ
- 43
@atulit_gaur ·
i think the fusion of physics and ai is the most beautiful one we train neural networks with optimization methods rooted in physics, diffusion models borrow ideas from thermodynamics, physics informed neural networks solve differential equations by embedding physical laws directly into learning, energy based models treat learning as the search for low energy states, statistical mechanics has shaped the mathematics behind entire classes of machine learning algorithms etc the universe spent 13.8 billion years writing its laws and today we are teaching machines to think by borrowing those very laws
- 44
@hugobowne ·
“You still use pull requests? I wouldn’t even do that anymore. Just push it straight to trunk, have your agent summarize it.” That’s @gregce10, co-founder and CPO of SpecStory. He previously worked at GitHub, Dropbox and Google, and was CPO at Pluralsight. And he kept going: - PRs are the limiting gate when agents produce more code than humans can review. - The model should never decide when its own work is finished. Put the deterministic checks somewhere it cannot access. - Git is probably here to stay. Whether GitHub remains the platform, “we’ll see.” @HanchungLee came at the same problem from the evaluation side. Han is Director of Machine Learning at Moody’s and works on SkillsBench, evaluating skills across combinations of models and agent harnesses. - An agent is the model plus its harness. You need to evaluate the complete system. - A green check proves nothing if the agent found a way to game the task. - Your agent could delete the failing test and declare success. Both are figuring out how to turn masses of agent-generated slop into signal. Greg mined 516 saved agent sessions to recover the decisions and intent behind the work, identify recurring practices, and forge the ones he approved into reusable skills. Han runs skills inside controlled environments, grades the result, and preserves the complete trajectory so we can inspect what the agent actually did. Preserve the intent. Inspect the trajectory. Verify the result. Turn what works into skills. Full episode in the replies 👇
Watch video
- 45
@DylanFeltus ·
curious about building your own model? the actual roadmap: step 1: understand transformers → Karpathy "Let's build GPT" https://t.co/QRC54Uo5es → "Attention Is All You Need" paper https://t.co/JEN33a8Bxw → HF LLM course https://t.co/PbZUWmJbWz → https://t.co/UGTmB0L8Z6 step 2: fine-tune a model (fastest way to "your own") → Unsloth https://t.co/LwOpqmmPz2 — runs on consumer hardware → Axolotl https://t.co/3Go38MIxxT → HF TRL https://t.co/4Wtg4RIQHO — SFT, DPO, GRPO step 3: train from scratch → nanochat https://t.co/ub1AvqQw77 — GPT-2 for ~$48 → LitGPT https://t.co/LajhHXxZnP → datasets: FineWeb, RedPajama, The Pile (all on HF) 📚 more resources: → "Deep Learning" by Goodfellow, Bengio & Courville — the textbook https://t.co/Zrr0jwiALu → "Build a Large Language Model (From Scratch)" by Sebastian Raschka — hands-on, code-first https://t.co/3OeGbGCycv → "Designing Machine Learning Systems" by Chip Huyen — the production/systems side → Karpathy's "Neural Networks: Zero to Hero" playlist https://t.co/iKLe8rtDso → 3Blue1Brown neural networks series https://t.co/knxwlyF9eq
- 46
@Meta_Engineers ·
Our Ranking Engineer Agent (REA) autonomously executes key steps across the end-to-end machine learning lifecycle for ads ranking models. REA reduces the need for manual intervention, managing asynchronous workflows spanning days to weeks through a hibernate-and-wake mechanism, with human oversight at key strategic decision points. Read this post that covers REA’s ML experimentation capabilities: autonomously generating hypotheses, launching training jobs, debugging failures, and iterating on results: https://t.co/W5AcATUFXX
- 47
- 48
@sabir_huss50540 ·
A team of two dozen researchers just trained a 35B model to improve the process of building AI, put it on a single consumer graphics card, and released the whole thing. The paper is called Frontis-MA1. It came out of Tsinghua and a lab called https://t.co/wlcZKlCube on July 30. Here's the part that should stop you. Their model, Frontis-MA1-35B, runs on one RTX 4090 capped at 12GB of VRAM. That's a gaming card. On MLE-Bench Lite, a benchmark where an agent has to actually build and iterate a real machine learning solution, it scores a 71.21% medal average. That beats GPT-5.5 with Codex. It approaches GPT-5.6 Sol and Kimi K3, a 2.8 trillion parameter model. 35 billion parameters, matching 2.8 trillion. On a card you can buy. The trick isn't a bigger model. It's a loop. Frontis-MA1 is trained on four atomic moves, Draft, Improve, Debug, and Crossover, the exact same moves it then runs at inference time to evolve its own solutions. The engine that runs the search was trained on the search. The model that proposes the next experiment is shaped by the results of the last one. They call it meta-evolution. On its Qwen base, post-training alone moved the score from 39.39% to 60.61%. The search loop pushed it to 71.21%. Training and search stacked. This is the phrase the authors are careful with. Recursive self-improvement. The thing every AI safety debate circles around. They do not claim they built it. They call this a concrete step toward it, a model that helps build AI by learning from watching itself build AI. Then they released everything. The weights. The training data. The sandbox. The search framework. Open, so anyone can run the loop. The frontier labs guard trillion-parameter models behind an API and call the automation of AI research their moat. A 35B model that fits on a graphics card just walked up to that line and open-sourced the map.
- 49
@TDataScience ·
"This time I found that making progress in machine learning, or any field, really, often comes down to the following five things. They are patience, discipline, optimism, good projects, and good teams." Pascal Janetzky reflects on 8.5 years as a machine learning practitioner. https://t.co/ruVt9OjS2p
- 50
Best Tweets by Topic