Sam Austinon November 7, 202616 min read

Awesome Lists for AI and ML: The Curated GitHub Collections Worth Bookmarking

Awesome Lists for AI and ML: The Curated GitHub Collections Worth Bookmarking
Contents

Figure 1: Awesome Lists for AI and ML — you google "best machine learning resources," and a thousand tabs later you still can't remember which list had that anomaly detection library.

You google "best machine learning resources," and a thousand tabs later you still can't remember which list had that anomaly detection library. Curated "awesome lists" promise to fix that: someone else already did the sorting. But GitHub now holds thousands of awesome lists, and half of them are forks, abandoned copies, or link graveyards.

I cannot claim I have read every list below end to end, so I will give you my honest analysis of what each one offers and how to judge it. I also searched for current details, and I could not verify recent commit activity for most lists, so check freshness yourself. IMO, the skill that matters isn't finding lists but vetting them in two minutes. Let us build that skill.

What Is an Awesome List?

An awesome list is a GitHub repository, usually a single README, that collects links on one topic: libraries, papers, courses, datasets, or tools. The original awesome project started the convention, and a flood of topic-specific copies followed. One meta-index describes the situation as 8,500+ awesome lists with over 1,319,000 curated resources. That is a lot of links, and nobody has vetted them all. Context Awesome

Good lists save you time. Bad lists send you to dead repos. Ever clicked a "best of" link and landed on a 404? Same pain, bigger scale.

How to Vet a List in Two Minutes

Run this checklist before you bookmark:

  • Freshness: check the last commit date, not the star count
  • Curation: does the list explain what makes an entry "awesome," or does it accept everything?
  • Descriptions: one-line summaries beat bare links
  • Structure: sections, tables of contents, and consistent formatting signal care
  • Contribution rules: a CONTRIBUTING file suggests active gatekeeping
  • Duplicates: several repos share the same name, so confirm you have got the canonical one
  • Link health: click ten random links and count the failures

Stars measure past popularity. Recent commits measure present usefulness.

Meta-Lists: Lists of Lists

When you do not know where to start, start here.

Awesome and Awesome-Awesomeness

One index describes the root awesome repo as awesome lists about all kinds of interesting topics, next to awesome-awesomeness, a curated list of awesome awesomeness, and similar meta-lists. They help you discover lists on topics you have not thought of. GitHub

Ultimate Awesome

The ultimate-awesome project tries to index every awesome list on every topic, including lists of awesome lists, updated daily. Use it as a search layer. Search your topic, then judge the results with the checklist above. GitHub

Awesome-Machine-Learning-Resources

This one appears in the same index as a curated list of awesome lists across all machine learning topics, organized by paradigm, task, application, model, ethics, datasets, frameworks, and tutorials. I would start here for broad ML coverage. GitHub

Aggregator Sites

Third-party sites index awesome lists and show metadata like stars, resource counts, and update times. That helps you compare lists quickly, but remember these sites also surface forks and duplicates. One browsing page I found listed several different repos all named awesome-mcp-servers, with very different resource counts and star counts. Always confirm the owner and activity before you trust a name. Context Awesome

Core AI and ML Lists

General Machine Learning and Deep Learning

Classic general lists include awesome-machine-learning (frameworks and libraries by language), awesome-deep-learning, and framework-focused lists like awesome-tensorflow and awesome-pytorch-list, a comprehensive list of PyTorch models, implementations, helper libraries, and tutorials. I am describing the first two from memory, so verify their current status before relying on them.

awesome-machine-learning | awesome-deep-learning | awesome-pytorch-list

Production ML and MLOps

awesome-production-machine-learning collects open-source libraries to deploy, monitor, version, and scale ML systems. If you have ever wondered which tool handles drift monitoring or feature stores, start there. GitHub

Large Language Models

Several lists track LLM resources. One meta-index lists awesome-llm, a curated list of large language model resources. Others focus on applications. One personal starred list shows awesome-llm-apps, a collection of 100+ AI agent and RAG apps you can run, clone, and customize, at roughly 112,000 stars, but treat the star count as a snapshot from someone's saved list. GitHub | awesome-llm-apps

Running Models Locally

The awesome-local-llms list stands out for its method. It compares open-source local LLM inference projects by popularity and activeness metrics. That is exactly the kind of evidence-based curation I want in a list. Context Awesome

Agents and MCP

The Model Context Protocol attracted its own cluster of lists. A personal starred list shows an awesome-mcp-servers collection at about 88,000 stars. Meanwhile, other repos with the same name appear in aggregator indexes with different owners and sizes. Because MCP servers can touch your data and systems, audit any server before you connect it, no matter how popular the list that recommended it.

Related lists include awesome-ai-market-maps, a list of 500+ AI market maps from 2026 and 2025, useful if you track the industry, and awesome-papers-autonomous-agent, a collection of recent papers on RL-based and LLM-based agents.

awesome-mcp-servers | awesome-ai-market-maps | awesome-papers-autonomous-agent

Lists for Reinforcement Learning

Since this series focuses on RL, here is what I found.

Awesome-RL and Its Cousins

The awesome-rl list organizes resources into codes, theory, lectures, books, surveys, papers, applications, game playing, robotics, control, operations research, and human-computer interaction. I like that it includes an operations research section. The catch: one aggregator snapshot shows the original as last synced seven months before that snapshot, and it appears under many forks. Another list, awesome-reinforcement-learning, covers frameworks, libraries, tools, environments, tutorials, and papers, and mentions older projects such as Dopamine and Acme. Verify whether those projects still receive updates. Ecosyste.ms

Awesome-Offline-RL

This list is a standout. It describes itself as a collection of research and review papers for offline reinforcement learning, maintained by Haruka Kiyohara and Yuta Saito. Its sections cover reviews and surveys, offline RL methods, off-policy evaluation and learning, benchmarks, applications, blogs, podcasts, workshops, and tutorials. One index shows about 1,078 stars. The off-policy evaluation section matters for anyone following this series' evaluation advice. GitHub

Freshness caution: I saw conflicting signals about when it last changed, with one snapshot showing a push date in early 2024 and aggregator pages showing recent syncs. Syncing is not the same as committing, so check the repo's commit history directly.

Explainable RL and Applications

The aggregator index also lists awesome-explainable-reinforcement-learning, a survey-style list of concepts, algorithms, and challenges. Application-specific lists show up too, such as a list of ML and deep learning content from the mobility and transportation industry. If you work on the problems from earlier articles, like dispatch and routing, search for lists in your own industry.

awesome-explainable-reinforcement-learning

Lists for Time Series, Edge, and Specialized Topics

  • Time series anomaly detection: the meta-index lists awesome-TS-anomaly-detection, a list of tools and datasets for anomaly detection on time-series data. That connects to the first article in this series. GitHub
  • Edge ML: awesome-edge-machine-learning collects research papers, inference engines, challenges, books, and meetups for edge deployment. Context Awesome
  • Tabular data: one list I found showcases CatBoost, which it calls the best framework for tabular data. That is the list author's opinion, so judge it yourself. Context Awesome
  • Practitioner collections: awesome-ai-data-github-repos gathers important GitHub repos for ML, AI, and data science practitioners. Context Awesome

Quick Comparison

Here is my fast take on where to start:

  • Awesome and ultimate-awesome: best for discovering lists
  • awesome-machine-learning-resources: best for broad ML coverage
  • awesome-production-machine-learning: best for MLOps tooling
  • awesome-local-llms: best for evidence-based local inference choices
  • awesome-mcp-servers variants: best for agent tooling, with caution
  • awesome-rl and awesome-reinforcement-learning: best for RL breadth
  • awesome-offline-rl: best for offline RL and OPE papers
  • awesome-TS-anomaly-detection: best for time series anomaly work

How to Build Your Own Bookmark System

Do not star fifty lists and call it research. Try this:

  1. Pick three lanes that match what you build this month
  2. Choose one list per lane using the vetting checklist
  3. Star those lists, and watch releases only if the list stays active
  4. Search the list before Google when you need a tool
  5. Test two options from the list instead of reading ten descriptions
  6. Contribute a fix when you find a dead link, which is the fastest way to learn what the maintainers value
  7. Prune quarterly: drop lists that stopped updating

A focused watchlist beats a starred junk drawer.

Common Mistakes to Avoid

  • Trusting stars over commits. A 90,000-star list that stopped updating in 2023 misleads.
  • Using forks by accident. Aggregator pages show many repos with identical names, so check the owner.
  • Treating inclusion as endorsement. A list entry does not mean someone tested the tool.
  • Connecting unvetted tools. MCP servers, agents, and scripts from a list can access your systems, so read the code first.
  • Hoarding without using. Bookmarks do not build skills, projects do.
  • Ignoring licenses. A list can point to repos you cannot use commercially.
  • Skipping the contribution rules. If you submit a link, follow the guidelines or maintainers will close your PR.

Did you notice that most mistakes involve trusting a signal that is easy to see over one that matters?

Final Thoughts

The best awesome list is the one someone still maintains and you still open. Start with a meta-list to discover options, pick one active list per lane, and vet each with the two-minute checklist. Treat stars as rumors, commit dates as evidence, and every linked tool as untrusted until you read it.

So pick one lane this week, like offline RL, local LLMs, or MLOps, and open its best list. Click ten links, count the dead ones, and either bookmark it or move on. If you find a broken link, fix it with a pull request. Congratulations: you just contributed to open source, and your first PR can be a one-line change.

Frequently Asked Questions

How do I know if an awesome list is still maintained?

Check the last commit date, not the star count. Look for commits within the last 30-60 days, recent releases, and active issue responses. Stars measure past popularity; commits measure present usefulness.

Which awesome list should I start with for machine learning?

Start with the meta-lists: awesome and ultimate-awesome for discovery, then awesome-machine-learning-resources for broad ML coverage. For specific needs, check awesome-production-machine-learning (MLOps) or awesome-local-llms (local inference).

Are stars a good indicator of list quality?

No. Stars measure past popularity, not current usefulness. A 90,000-star list that stopped updating in 2023 is worse than a 2,000-star list updated last week. Check commit dates, not stars.

How do I avoid using forked or outdated lists?

Check the repo owner against the canonical owner. Aggregator sites often show multiple repos with the same name (e.g., awesome-mcp-servers). Always verify the owner URL and check recent commit activity.

Should I star every list I find useful?

No. Star the 1-2 best lists per topic lane you actually use. A focused watchlist beats a starred junk drawer. Contribute fixes instead of hoarding stars.

Sources

What are You Looking For?

esc