Projects Recently Starred by Experts

Starred 15 hours ago by

Piero Molino(Cofounder of Predibase).

lucebox-hub by Luce-Org

46.0%

273

Optimized LLM inference for specific hardware

Hand-tuned CUDA kernels and fused passes for maximum throughput and energy efficiency (tok/J).
Implements advanced techniques like speculative decoding (DFlash) and hybrid Attention/DeltaNet architectures.
Tailored software stack for consumer GPUs, enabling large context windows and high token generation rates.
Achieves state-of-the-art performance per watt by rewriting software for specific silicon.

Created 2 weeks ago

Updated 2 days ago

Starred 16 hours ago by

Pietro Schirano(Founder of MagicPath). Also starred by

Victor Taelin(Author of Bend, Kind, HVM).

rvllm by m0at

24.8%

524

High-performance LLM inference engine in Rust

Drop-in replacement for vLLM, offering dramatically improved resource efficiency and faster startup times.
Leverages Rust's performance, JIT-compiled fused CUDA kernels, and FP8 inference for optimized throughput and reduced memory footprint.
Features advanced attention mechanisms (FA3 v3), CUDA graph replay, and cuBLAS autotuning for low-latency, high-concurrency serving.
Achieves a 31x smaller binary size and 20x faster cold starts compared to Python-based alternatives.

Created 3 weeks ago

Updated 14 hours ago

Starred 18 hours ago by

Kevin Hou(Head of Product Engineering at Windsurf). Also starred by

Chuan Li(Chief Scientific Officer at Lambda),

Enes Akar(Cofounder of Upstash),

Shyamal Anadkat(Research Scientist at OpenAI), and

13 more.

gstack by garrytan

8.1%

75k

Specialized AI coding assistant framework

Transforms generic AI into a team of specialized agents for planning, review, shipping, and QA.
Leverages browser automation and structured reporting for end-to-end QA and live URL inspection.
Enables parallel agent sessions and context-aware mode switching for complex software development workflows.
Integrates cookie import for authenticated testing and generates data-driven engineering retrospectives.

Created 1 month ago

Updated 17 hours ago

Starred 20 hours ago by

Paul Stamatiou(Cofounder of Limitless). Also starred by

Wes McKinney(Author of Pandas),

Ed Huang(Cofounder of PingCAP),

Tobi Lutke(Cofounder of Shopify), and

1 more.

gbrain by garrytan

47.2%

AI agent knowledge backbone for personal data

Manages personal markdown knowledge bases, acting as an AI agent's persistent memory.
Enables RAG with hybrid search (keyword + vector) over personal documents.
Features a compiled truth + timeline model for structured, evidence-based knowledge.
Offers CLI, SDK, and MCP server for seamless agent integration.

Created 2 weeks ago

Updated 14 hours ago

Starred 1 day ago by

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

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

Tim J. Baek(Founder of Open WebUI),

Julien Chaumond(Cofounder of Hugging Face), and

1 more.

LlamaBarn by ggml-org

3.3%

Easy local LLM deployment on Mac

Run diverse LLMs with one click using a native macOS menu bar app.
Features a chat UI and REST API, built upon the llama.cpp inference engine.
Automatic hardware configuration ensures optimal performance and stability for local models.
Tiny (~12MB) app with curated catalog, simplifying LLM access without system-wide installs.

Created 9 months ago

Updated 2 days ago

Starred 1 day ago by

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

Travis Fischer(Founder of Agentic).

hyperframes by heygen-com

73857.1%

HTML-native video rendering for AI agents

Create dynamic videos from HTML, text, or data via AI coding assistants.
Integrates GSAP animations, shader transitions, and data visualizations.
Enables deterministic, automated video pipelines with an HTML-first approach.
AI agent skills simplify composition authoring, iteration, and rendering.

Created 1 month ago

Updated 19 hours ago

Starred 1 day ago by

Joe Walnes(Head of Experimental Projects at Stripe). Also starred by

Gregor Zunic(Cofounder of Browser Use),

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

Jeffrey Morgan(Cofounder of Ollama), and

22 more.

ComfyUI by Comfy-Org

0.7%

109k

Visual AI engine for diffusion models, API, and backend

Node-based interface creates complex Stable Diffusion workflows without coding.
Supports SD1.x, SD2.x, SDXL, Stable Video Diffusion, SD3, and more.
Optimizes memory management for low VRAM GPUs; standalone VAEs/CLIP models.
Loads workflows from PNG/WebP/FLAC; saves/loads as JSON; works offline.

Created 3 years ago

Updated 20 hours ago

Starred 1 day ago by

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

Yaowei Zheng(Author of LLaMA-Factory).

EverOS by EverMind-AI

9.7%

Enterprise-grade AI memory system for persistent understanding

Enables AI to build understanding from past conversations, creating coherent narratives and living user profiles.
Features evidence-based perception with hybrid retrieval (RRF, semantic, keyword) and LLM-guided agentic recall.
Leverages multi-level memory structures (MemCells, episodes) and integrates with MongoDB, Elasticsearch, and Milvus.
Achieved 92.3% reasoning accuracy on the LoCoMo benchmark, demonstrating advanced AI understanding.

Created 5 months ago

Updated 16 hours ago

Starred 2 days ago by

Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"). Also starred by

Max Liu(Cofounder of PingCAP),

Thomas Wolf(Cofounder of Hugging Face), and

Pawel Garbacki(Cofounder of Fireworks AI).

ShowUI by showlab

1.4%

Vision-language-action model for GUI agent & computer use (CVPR 2025 paper)

End-to-end model takes visual and language inputs, outputs actions for GUI interaction.
Supports Qwen2.5-VL, vllm inference, int8 quantization, and API calling via Gradio.
Training codebase supports DeepSpeed, BF16, QLoRA, SDPA, FlashAttention2, Liger-Kernel.
Supports navigation tasks: Mind2Web, AITW, Miniwob training and evaluation.

Created 1 year ago

Updated 2 months ago

Starred 2 days ago by

Eric Zhang(Founding Engineer at Modal). Also starred by

Chaoyu Yang(Founder of Bento),

Jeff Hammerbacher(Cofounder of Cloudera),

Vincent Weisser(Cofounder of Prime Intellect), and

1 more.

InferenceX by SemiAnalysisAI

4.8%

827

Real-time LLM inference performance benchmarking

Automated, continuous benchmarking of open-source inference frameworks and models.
Tracks LLM token throughput and performance across diverse hardware/software.
Highlights kernel optimizations, distributed strategies, and scheduling innovations.
Provides live indicators of performance per dollar and tokens per Megawatt.

Created 9 months ago

Updated 1 day ago

...