Projects Recently Starred by Experts

Starred 22 hours ago by

Didier Lopes(Founder of OpenBB). Also starred by

Philipp Schmid(DevRel at Google DeepMind) and

Travis Fischer(Founder of Agentic).

offmute by SouthBridgeAI

0.5%

567

CLI tool for meeting transcription and analysis using Gemini models

Transcribes audio/video, identifying speakers; diarization via LLM.
Generates meeting reports: key points, action items, participant profiles.
Supports multiple processing tiers, balancing cost/quality.
Employs "Spreadfill" technique for coherent, detailed report generation.

Created 1 year ago

Updated 2 months ago

Starred 23 hours ago by

Didier Lopes(Founder of OpenBB). Also starred by

Sebastian Raschka(Author of "Build a Large Language Model (From Scratch)"),

Peter Norvig(Author of "Artificial Intelligence: A Modern Approach"; Research Director at Google),

Gabriel Almeida(Cofounder of Langflow), and

13 more.

LLMs-from-scratch by rasbt

0.3%

97k

Educational resource for LLM construction in PyTorch

GPT-like LLM implementation from scratch, step-by-step.
Covers pretraining and finetuning techniques.
Includes code for loading pretrained weights.
Contains bonus material on PEFT with LoRA.

Created 2 years ago

Updated 2 weeks ago

Starred 1 day ago by

Jeremy Howard(Cofounder of fast.ai). Also starred by

Didier Lopes(Founder of OpenBB),

Luis Capelo(Cofounder of Lightning AI),

Jiaming Song(Chief Scientist at Luma AI), and

8 more.

slime by THUDM

5.8%

LLM post-training framework for RL scaling

Enables efficient training via Megatron integration with SGLang.
Supports flexible data generation workflows via custom interfaces.
Includes HF checkpoint conversion to Megatron's torch_dist format.

Created 1 year ago

Updated 12 hours ago

Starred 1 day ago by

Elvis Saravia(Founder of DAIR.AI). Also starred by

Michael Chiang(Cofounder of Ollama),

Andy Konwinski (Cofounder of Perplexity, Databricks),

Jeff Hammerbacher(Cofounder of Cloudera), and

3 more.

omnigent by omnigent-ai

259.4%

Unify and manage AI agents across platforms

Supervise multiple AI agents (Claude, Codex, Pi, custom) in a single session.
Enables cross-device collaboration and real-time session syncing from terminal to phone.
Supports agent governance via policies for sandboxing, spend caps, and action approvals.
Allows defining custom agents via YAML, including tools and sub-agent delegation.

Created 1 week ago

Updated 13 hours ago

Starred 2 days ago by

Didier Lopes(Founder of OpenBB). Also starred by

Luis Capelo(Cofounder of Lightning AI).

triattention by WeianMao

2.2%

791

LLM long reasoning acceleration via KV cache compression

Compresses KV cache by 10.7x, enabling efficient long context reasoning.
Achieves 2.5x throughput boost with no accuracy loss on complex tasks.
Uses trigonometric frequency-domain compression and integrates with vLLM/OpenCLaw.
Facilitates local deployment on memory-constrained GPUs like RTX 4090.

Created 2 months ago

Updated 3 days ago

Starred 2 days ago by

Bert Belder(Coauthor of Deno).

clawpatrol by denoland

4.8%

853

Security firewall for AI agents

Gates agent actions against custom HCL rules, preventing destructive operations like unauthorized SQL or Kubernetes commands.
Parses wire-level traffic across protocols (SQL, K8s, HTTP) using CEL expressions for granular control.
Offers flexible deployment: standalone gateway proxy, host-wide WireGuard tunnel, or per-process agent traffic tunneling.
Enables human approval gates for sensitive agent actions before they reach production systems.

Created 1 month ago

Updated 1 day ago

Starred 2 days ago by

Didier Lopes(Founder of OpenBB). Also starred by

Wei-Lin Chiang(Cofounder of LMArena) and

Jeff Hammerbacher(Cofounder of Cloudera).

GLM-5 by zai-org

40.8%

Large language model for complex agentic systems

Powers complex systems engineering and long-horizon agentic tasks with advanced reasoning and planning.
Leverages DeepSeek Sparse Attention for efficient long-context handling and reduced deployment costs.
Utilizes novel asynchronous RL infrastructure ('slime') for enhanced training efficiency and fine-grained post-training.
Achieves top open-source performance in coding, reasoning, and long-term operational agentic benchmarks.

Created 4 months ago

Updated 1 day ago

Starred 2 days ago by

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

Zhiqiang Xie(Coauthor of SGLang) and

Ying Sheng(Coauthor of SGLang).

TileRT by tile-ai

5.3%

Ultra-low-latency LLM inference runtime

Delivers extreme responsiveness for interactive AI, agents, and real-time applications.
Utilizes a tile-level compiler engine for fine-grained task scheduling and overlap.
Achieves millisecond-level TPOT for large models on multi-GPU systems.
Maximizes hardware utilization by overlapping compute, I/O, and communication.

Created 7 months ago

Updated 1 week ago

Starred 3 days ago by

Jason Knight(Director AI Compilers at NVIDIA; Cofounder of OctoML).

pi-acp by svkozak

3.3%

461

ACP adapter for AI coding agents

Bridges pi coding agent with ACP clients (e.g., Zed editor) via JSON-RPC 2.0.
Streams assistant output, maps tool calls, and surfaces code locations/diffs in the editor.
Integrates file-based and built-in slash commands for enhanced agent interaction.
Supports session persistence and history, enabling resumed coding workflows.

Created 6 months ago

Updated 3 days ago

Starred 3 days ago by

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

Jason Knight(Director AI Compilers at NVIDIA; Cofounder of OctoML) and

Mehdi Amini(Author of MLIR; Distinguished Engineer at NVIDIA).

cutile-rs by NVlabs

81.2%

617

Safe, tile-based GPU kernel programming in Rust

Write GPU kernels in Rust with a safe, tile-based DSL and asynchronous execution.
Pure Rust compiler pipeline and host-side API for efficient tensor operations.
JIT-compiles CUDA kernels, supporting NVIDIA GPUs with compute capability sm_80+.
Research project exploring advanced GPU programming paradigms within the Rust ecosystem.

Created 3 months ago

Updated 2 days ago

...