Projects Recently Starred by Experts

Starred 6 hours ago by

Didier Lopes(Founder of OpenBB). Also starred by

Jason Huggins(Creator of Selenium),

Luis Capelo(Cofounder of Lightning AI),

Taranjeet Singh(Cofounder of Mem0), and

16 more.

hermes-agent by NousResearch

6.8%

137k

Autonomous AI agent for persistent learning and task automation

Integrates across messaging platforms (Telegram, Discord, Slack, WhatsApp) with scheduled automations.
Supports diverse LLM providers, voice transcription/synthesis, and web browsing tools.
Offers robust sandboxing (Docker, SSH, Modal) and a skill-based knowledge system.
Research-ready for parallel batch processing and reinforcement learning model training.

Created 9 months ago

Updated 6 hours ago

Starred 8 hours ago by

Li Jiang(Coauthor of AutoGen; Engineer at Microsoft). Also starred by

Tim Suchanek(Founder of expand.ai),

Jeff Hammerbacher(Cofounder of Cloudera),

Travis Fischer(Founder of Agentic), and

2 more.

agent-skills by addyosmani

20.6%

32k

Production-grade engineering workflows for AI coding agents

Standardizes software development phases (Define, Plan, Build, Verify, Review, Ship) with structured, verifiable skills.
Integrates with AI agents via Markdown skills, enforcing best practices and quality gates.
Embeds Google engineering principles like Hyrum's Law, test pyramid, and Shift Left into agent workflows.
Skills include anti-rationalization tables and non-negotiable verification steps for robust agent output.

Created 2 months ago

Updated 1 day ago

Starred 9 hours ago by

Wes McKinney(Author of Pandas). Also starred by

Georgios Konstantopoulos(CTO, General Partner at Paradigm),

Matt McClure(Cofounder of Mux), and

Amir Salihefendic(Founder of Doist).

hunk by modem-dev

433.9%

Terminal diff viewer for agentic code review

Interactive, review-first diffing for agent-authored code changesets.
Integrates inline AI annotations and agent comments directly within code views.
Features multi-file review streams, responsive layouts, and watch mode for dynamic updates.
Enables remote control of live diff sessions and loading agent context from files.

Created 1 month ago

Updated 7 hours ago

Starred 11 hours ago by

Georgios Konstantopoulos(CTO, General Partner at Paradigm). Also starred by

Andrey Vasnetsov(Cofounder of Qdrant),

Wing Lian(Founder of Axolotl AI),

Patrick von Platen(Author of Hugging Face Diffusers; Research Engineer at Mistral), and

2 more.

rtk by rtk-ai

11.2%

44k

CLI proxy for massive LLM token reduction

Compresses CLI command outputs for LLMs, slashing token consumption by 60-90%.
Integrates via hooks to transparently rewrite commands, optimizing LLM context.
High-performance Rust binary provides detailed token savings analytics and discovery.
Optimizes common dev commands (git, grep, files) with full output recovery.

Created 3 months ago

Updated 1 hour ago

Starred 13 hours ago by

Pawel Garbacki(Cofounder of Fireworks AI). Also starred by

Max Howell(Author of Homebrew),

Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"),

Amin Ahmad(Cofounder of Vectara), and

30 more.

litellm by BerriAI

1.5%

46k

SDK/proxy for calling 100+ LLM APIs using the OpenAI format

Standardizes LLM API calls: completion, embedding, image generation.
Supports 30+ providers: OpenAI, Azure, Bedrock, HuggingFace, VertexAI, Groq.
Offers proxy server with auth, logging, cost tracking, rate limiting.
Provides async completion, streaming, and observability callbacks.

Created 2 years ago

Updated 8 hours ago

Starred 15 hours ago by

Luis Capelo(Cofounder of Lightning AI). Also starred by

Junyang Lin(Core Maintainer at Alibaba Qwen),

Shizhe Diao(Author of LMFlow; Research Scientist at NVIDIA),

Meng Zhang(Cofounder of TabbyML), and

7 more.

miles by radixark

5.4%

Enterprise RL for large-scale MoE models

Production-grade reinforcement learning framework for massive Mixture-of-Experts models.
Achieves stable inference alignment via true on-policy training and numerical consistency.
Enhances memory robustness and accelerates rollouts using speculative training methods.
Supports new hardware and targets multimodal training for advanced AI applications.

Created 7 months ago

Updated 2 hours ago

Starred 17 hours ago by

Luis Capelo(Cofounder of Lightning AI). Also starred by

Jeff Hammerbacher(Cofounder of Cloudera),

Wing Lian(Founder of Axolotl AI), and

George Hotz(Author of tinygrad; Founder of the tiny corp, comma.ai).

TileKernels by deepseek-ai

6.3%

Optimized GPU kernels for LLM operations

Efficient Mixture of Experts (MoE) routing and gating for LLM training and inference.
Advanced FP8/FP4/E5M6 quantization with fused operations for memory and compute gains.
TileLang-based kernels achieve near hardware limits for compute intensity and memory bandwidth.
Supports novel architectures via Engram gating and Manifold HyperConnection (mHC) kernels.

Created 2 weeks ago

Updated 2 weeks ago

Starred 17 hours ago by

Luis Capelo(Cofounder of Lightning AI). Also starred by

Lysandre Debut(Chief Open-Source Officer at Hugging Face).

sglang-omni by sgl-project

12.0%

261

High-performance framework for multi-modal, multi-stage AI models

Orchestrates complex multi-stage pipelines for omni models with multi-modal inputs/outputs.
Supports multi-modal AI, extending LLM-specific architectures to handle diverse data types.
Leverages SGLang integration for optimized performance and real-time API serving.
Provides an OpenAI-compatible server for seamless integration and deployment.

Created 4 months ago

Updated 4 hours ago

Starred 21 hours ago by

Wing Lian(Founder of Axolotl AI). Also starred by

Chaoyu Yang(Founder of Bento),

Shizhe Diao(Author of LMFlow; Research Scientist at NVIDIA),

Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), and

6 more.

openevolve by algorithmicsuperintelligence

0.9%

Coding agent for scientific/algorithmic discovery, based on AlphaEvolve paper

Evolves entire code files using LLMs, supports multiple languages.
Employs LLM ensemble for code generation, guided by program database.
Uses asynchronous pipeline for high-throughput candidate evaluation.
Includes examples for circle packing and function minimization tasks.

Created 11 months ago

Updated 1 month ago

Starred 1 day ago by

Thomas Dohmke(Former CEO of GitHub). Also starred by

Elie Bursztein(Cybersecurity Lead at Google DeepMind) and

Wes McKinney(Author of Pandas).

GitNexus by abhigyanpatwari

7.6%

37k

Browser-based code intelligence engine

Client-side code analysis engine builds interactive knowledge graphs for deep code understanding.
Full WASM stack: Tree-sitter parsing, KuzuDB graph/vector DB, and local embeddings run in-browser.
Hybrid search (BM25, semantic, graph expansion) powers a Graph RAG agent with grounded citations.
Zero-server architecture ensures code privacy and zero deployment cost.

Created 9 months ago

Updated 2 hours ago

...