M. Yusuf Sarıgöz

exercises-dataset. 1,324-exercise fitness dataset — animation GIFs, 180×180 thumbnails, muscle-group & equipment data, and step-by-step instructions in 6 languages. The exercise data layer behind the LogPress app.

OScaR-KV-Quant. 🏆 OScaR: The Occam's Razor for Extreme KV Cache Quantization in LLMs and Beyond — redefining the accuracy-efficiency Pareto front for X-LLMs KV quantization.

PLFM_RADAR. Open-source, low-cost 10.5 GHz PLFM phased array RADAR system

science-skills. GDM Science Skills to speed up agentic scientific workflows with better grounding and higher token efficiency. Integrate insights from AlphaGenome, AFDB, UniProt and 30+ other databases and tools.

semble. Fast and Accurate Code Search for Agents. Uses ~98% fewer tokens than grep+read

cq. An open standard for shared agent learning. Agents persist, share, and query collective knowledge so they stop rediscovering the same failures independently.

arxiv2md. Convert arxiv papers to markdown

Ctx2Skill. Code for "From Context to Skills: Can Language Models Learn from Context Skillfully? "

CodeWhale. Open-source, community-driven agent harness

nanochat.JAX. A legible and hackable port of Nanochat in JAX/Flax

RoundPipe. Large DNNs training framework for consumer GPUs

Nucleus. A high-performance, minimalistic container engine in Rust.

isanagent. Like Openclaw, but for ML engineering. Your always-on and autonomous ML engineer that works 24/7 to conduct research, generate datasets, train models, learn and evolve. Works on Colab, local GPU or over SSH. Zero infra needed

dirac. Coding Agent singularly focused efficiency and context curation. Reduces API costs by 50-80% vs other agent AND improves the code quality at the same time. Uses Hash Anchored edits, massively parallel operations, AST manipulation and many many other optimizations. https://dirac.run/

npcforge. Branching NPC dialogue generator with Yarn Spinner export, powered by afterimage.

agent-vault. A HTTP credential proxy and vault for AI agents like Claude Code, OpenClaw, Hermes, custom agents + harnesses, and more.

TIDE. Dynamic per-token early exit for LLM inference. Skip layers tokens don't need

smolvm. Portable, lightweight, self-contained virtual machines.

afterimage. Generate conversational, tool-calling, structured-output, and preference datasets — easily and at scale

VoxCPM. VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning

awesome-design-md. A collection of DESIGN.md files analysis by popular brand design systems. Drop one into your project and let coding agents generate a matching UI.

claw-code. An agent-managed museum exhibit, built in Rust with Gajae-Code / LazyCodex — developed and maintained with no human intervention.

turboquant-pytorch. From-scratch PyTorch implementation of Google's TurboQuant (ICLR 2026) for LLM KV cache compression. 5x compression at 3-bit with 99.5% attention fidelity.

email-server. Self-hosted personal email server with IMAP, SMTP, CalDAV/CardDAV. Privacy-focused alternative to Gmail/Outlook. Features: auto-TLS, DKIM/SPF/DMARC, greylisting, Prometheus metrics, audit logging.

email-explorer. A self-hosted, serverless worker for sending and receiving emails on Cloudflare

pageindex-rs. Hierarchical document indexing for RAG pipelines, written in Rust with Python bindings

FireRedVAD. A SOTA Industrial-Grade Voice Activity Detection & Audio Event Detection, supporting 100+ languages, outperforming Silero-VAD, TEN-VAD, FunASR-VAD and WebRTC-VAD

pdf_oxide. The fastest PDF library for Python and Rust. Text extraction, image extraction, markdown conversion, PDF creation & editing. 0.8ms mean, 5× faster than industry leaders, 100% pass rate on 3,830 PDFs. MIT/Apache-2.0.

ZeroFS. ZeroFS: A log-structured filesystem for S3. ZeroFS serves S3-compatible buckets as POSIX filesystems over NFS and 9P, or as raw block devices over NBD.

CUDA-Agent. CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation

turkicnlp. NLP Toolkit for Turkic Languages

OasisNvr. A high-performance Network Video Recorder written in Rust. Records RTSP streams from multiple IP cameras into pre-allocated binary pool files using a ring buffer strategy. Designed for HDD-based storage with sequential I/O optimization — no SSD required.

moonshine. Very low latency speech to text, intent recognition, and text to speech, for building voice agents and interfaces

RuVector. RuVector is a High Performance, Real-Time, Self-Learning Ai, Vector GNN, Memory DB built in Rust.

zeroclaw. Fast, small, and fully autonomous AI personal assistant infrastructure, any OS, any platform — deploy anywhere, swap anything 🦀

agentfs. The filesystem for agents.

functiongemma-afterimage-demo. Jupyter Notebook

Movie-Streaming-Web-Interface. A static movie browsing web interface built with HTML5 and CSS3, featuring a structured navigation system, sidebar icons, featured content area, categorized movie lists, and interactive visual styling designed to simulate a modern streaming platform layout.

iris.c. Flux 2 image generation model pure C inference

training-doctor. Python

rxiv-maker. Rxiv-Maker is a GitHub-native framework that converts markdown into a publication-ready PDF. It automates scientific preprint generation with a focus on reproducibility and version control

manuscript-rxiv-maker. Rxiv-Maker preprint source code

soprano-factory. Soprano-Factory: Train your own 2000x realtime text-to-speech model

mHC.cu. mHC kernels implemented in CUDA

SoE. [EACL'26] Repository for the Paper "Continual Pretraining on Encrypted Synthetic Data for Privacy-Preserving LLMs"

Chronicals. High-Performance LLM Fine-Tuning Framework - 3.51x faster than Unsloth

trailbase. An open, sub-millisecond, single-executable Firebase alternative with type-safe APIs, built-in WebAssembly runtime, realtime subscriptions, auth, and admin UI built on Rust, SQLite & Wasmtime.

aichat. All-in-one LLM CLI tool featuring Shell Assistant, Chat-REPL, RAG, AI Tools & Agents, with access to OpenAI, Claude, Gemini, Ollama, Groq, and more.

mistral-vibe. Minimal CLI coding agent by Mistral

semantic-soft-bootstrapping. A self-distillation based training method for long context reasoning in a single LLM without reinforcement learning

cutile-python. cuTile is a programming model for writing parallel kernels for NVIDIA GPUs

CUDA-L2. CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning

general-agentic-memory. A general memory system for agents, powered by deep-research

mini-kvstore-v2. Single node key–value store in Rust — segmented log, in-memory index, checksums, and manual compaction.

AsymmetryLoRA. Preprint: Asymmetry in Low-Rank Adapters of Foundation Models

dllm. dLLM: Simple Diffusion Language Modeling

pyversity. Fast Diversification for Search & Retrieval

helion. A Python-embedded DSL that makes it easy to write fast, scalable ML kernels with minimal boilerplate.

calm. Official implementation of "Continuous Autoregressive Language Models"

llink. Training and Inference for "Languages are Modalities: Cross-Lingual Alignment via Encoder Injection"

pynife. Nearly Inference Free Embeddings: make your RAG queries 500x faster

tokenmonster. Ungreedy subword tokenizer and vocabulary trainer for Python, Go, C++ & Javascript

nano-vllm. Nano vLLM

modded-nanogpt. NanoGPT (124M) in 90 seconds

DistilledPRAG. Python

triton_bwd. Automatic differentiation for Triton Kernels

LightMem. [ICLR 2026] LightMem: Lightweight and Efficient Memory-Augmented Generation

vlm-gym. RL gym for vision language models written in JAX

Math-VR-CodePlot-CoT. Math-VR Benchmark & CodePlot-CoT: Mathematical Visual Reasoning by Thinking with Code-Driven Images

mini-vec2vec. Jupyter Notebook

nanochat. The best ChatGPT that $100 can buy.

TinyRecursiveModels. Python

Paper2Video. Automatic Video Generation from Scientific Papers

Talk2FastVLM. Flask app to interact with FastVLM using CPU only

tunix. A Lightweight LLM Post-Training Library

DocLayout-YOLO. DocLayout-YOLO: Enhancing Document Layout Analysis through Diverse Synthetic Data and Global-to-Local Adaptive Perception

kuzu. Embedded property graph database built for speed. Vector search and full-text search built in. Implements Cypher.

ydl-rs. Rust

nano-gpt-oss. Learn the building blocks of how to build gpt-oss from scratch

nano-wal. A simple, lightweight Write-Ahead Log (WAL) implementation for append-only operations

hallbayes. Python

HRM. Hierarchical Reasoning Model Official Release

smoldb. A smol distributed database implemented from scratch 🐣

trackio. A lightweight, local-first, and free experiment tracking library from Hugging Face 🤗

vui. Real-time voice assistant — WebRTC streaming, faster-whisper ASR, local LLM, Vui Nano (300M) TTS. OpenAI Realtime API compatible. Voice cloning, barge-in, ~9× realtime on a 4090. Apache 2.0.

lemonade. Lemonade helps users discover and run local AI apps by serving optimized LLMs right from their own GPUs and NPUs. Join our discord: https://discord.gg/5xXzkMu8Zk

multimodal-research-agent. Python

gemma-mcp. MCP Client for Gemma-3

TTSizer. 🎙️ Automatically transcribe audio/video into high-quality, speaker-specific Text-To-Speech datasets

memvid. Memory layer for AI Agents. Replace complex RAG pipelines with a serverless, single-file memory layer. Give your agents instant retrieval and long-term memory.

gpt-fast. Simple and efficient pytorch-native transformer text generation in <1000 LOC of python.

circuit-tracer. Python

rules-for-ai. Tutorials are for humans, rules are for AI!

sie. Open-source inference server and production cluster for all the models your agent needs.

openevolve. Open-source implementation of AlphaEvolve

train-llm-from-scratch. A straightforward method for training your LLM, from downloading data to generating text.

ForkUnion. Lower-latency OpenMP-style minimalistic scoped thread-pool designed for 'Fork-Join' parallelism in Rust and C++, avoiding memory allocations, mutexes, CAS-primitives, and false-sharing on the hot path 🍴

gen-ai-livestream. Jupyter Notebook

nanoDeepResearch. A Deep Research agent from scratch

browserbee. 🐝 AI-powered browser assistant ("Cline for web browsing")

codeflash. Coding super-intelligence to find the most optimized Python code. Use it to optimize existing codebases or new Pull requests as a GitHub Action or a VS Code Extension.

llm-food. Serving files for hungry LLMs

toydb. Distributed SQL database in Rust, written as an educational project

Paper2Code. Paper2Code: Automating Code Generation from Scientific Papers in Machine Learning

EvoTune. Efficiently discovering algorithms via LLMs with evolutionary search and reinforcement learning.

opencode. A powerful AI coding agent. Built for the terminal.

dia. A TTS model capable of generating ultra-realistic dialogue in one pass.

flaxattention. Python

cocoindex. Incremental engine for long horizon agents 🌟 Star if you like it!

DFloat11. DFloat11 [NeurIPS '25]: Lossless Compression of LLMs and DiTs for Efficient GPU Inference

scaling-book. Home for "How To Scale Your Model", a short blog-style textbook about scaling LLMs on TPUs

Youtube_demos. All the demos in my Youtube Channel: Yeyu Lab

GenAI_Agents. 50+ tutorials and implementations for Generative AI Agent techniques, from basic conversational bots to complex multi-agent systems.

flashT5. A fast implementation of T5/UL2 in PyTorch using Flash Attention

diffusion-vs-ar. [ICLR 2025] Code for the paper "Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planning"

pydoll. Pydoll is a library for automating chromium-based browsers without a WebDriver, offering realistic interactions.

ferrules. Modern, fast, document parser written in 🦀

OpenManus. No fortress, purely open ground. OpenManus is Coming.

CommonLoopUtils. CLU lets you write beautiful training loops in JAX.

jaxgarden. A collection of reusable, high-performance, well-documented, thorough-tested layers and models in Jax

NeuralFlow. Visualize the intermediate output of Mistral 7B

DiffuLLaMA. [ICLR2025] DiffuGPT and DiffuLLaMA: Scaling Diffusion Language Models via Adaptation from Autoregressive Models

olmocr. Toolkit for linearizing PDFs for LLM datasets/training

vibe_coder. Voice assistant for VS Code

SMDM. Official PyTorch implementation for ICLR2025 paper "Scaling up Masked Diffusion Models on Text"

ggwave. Tiny data-over-sound library

gibberlink. Two conversational AI agents switching from English to sound-level protocol after confirming they are both AI agents

aibrix. Cost-efficient and pluggable Infrastructure components for GenAI inference

open-gemini-deep-research. Python

Moonlight. Muon is Scalable for LLM Training

imscore. Minimal Differentiable Image Reward Functions

GOT-OCR2.0. Official code implementation of General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model

LLaDA. Official PyTorch implementation for "Large Language Diffusion Models"

apprise. Apprise - Push Notifications that work with just about every platform!

audiobox-aesthetics. Unified automatic quality assessment for speech, music, and sound.

ktransformers. A Flexible Framework for Experiencing Heterogeneous LLM Inference/Fine-tune Optimizations

mlstm_kernels. Tiled Flash Linear Attention library for fast and efficient mLSTM Kernels.

deep-research-py. save 200 a month and use deep research right in your terminal. - port of https://github.com/dzhng/deep-research but in python

potpie. Context Graph for AI Native SDLC

production-stack. vLLM’s reference system for K8S-native cluster-wide deployment with community-driven performance optimization

open-tasks. Let LLMs schedule tasks for you and itself

open-thoughts. Fully open data curation for reasoning models

open-r1. Fully open reproduction of DeepSeek-R1

NumKong. SIMD-accelerated distances, dot products, matrix ops, geospatial & geometric kernels for 16 numeric types — from 6-bit floats to 64-bit complex — across x86, Arm, RISC-V, and WASM, with bindings for Python, Rust, C, C++, Swift, JS, and Go 📐

less_slow.py. Playing around "Less Slow" coding practices in Python, from numerical micro-kernels to coroutines, ranges, and polymorphic state machines

liburing. Library providing helpers for the Linux kernel io_uring support

titans-pytorch. Unofficial implementation of Titans, SOTA memory for transformers, in Pytorch

datafusion. Apache DataFusion SQL Query Engine

gline-rs. Inference engine for GLiNER models, in Rust

waste-oil-collection-management. JavaScript

taichi. Productive, portable, and performant GPU programming in Python.

concurrentqueue. A fast multi-producer, multi-consumer lock-free concurrent queue for C++11

letsearch. A vector DB so easy, even your grandparents can build a RAG system 😁

letsearch-client. Python client for letsearch — vector DB so easy, even your grandparents can build a RAG system 😁

reqwest. An easy and powerful Rust HTTP Client

Freeze-Omni. ✨✨Freeze-Omni: A Smart and Low Latency Speech-to-speech Dialogue Model with Frozen LLM

Liger-Kernel. Efficient Triton Kernels for LLM Training

codon. A high-performance, zero-overhead, extensible Python compiler with built-in NumPy support

podcastfy. An Open Source Python alternative to NotebookLM's podcast feature: Transforming Multimodal Content into Captivating Multilingual Audio Conversations with GenAI

extractous. Fast and efficient unstructured data extraction. Written in Rust with bindings for many languages.

vector. A high-performance observability data pipeline.

TokenFormer. [ICLR2025 Spotlight🔥] Official Implementation of TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters

OuteTTS. Interface for OuteTTS models.

FreeVC. FreeVC: Towards High-Quality Text-Free One-Shot Voice Conversion

gpu.cpp. A lightweight library for portable low-level GPU computation using WebGPU.

neural-speed. An innovative library for efficient LLM inference via low-bit quantization

unet.cu. UNet diffusion model in pure CUDA

llm.c. LLM training in simple, raw C/CUDA

Bend. A massively parallel, high-level programming language

Medical-RAG. Jupyter Notebook

medical-qa. Jupyter Notebook

quiet-star. Code for Quiet-STaR

burn. Burn is a next generation tensor library and Deep Learning Framework that doesn't compromise on flexibility, efficiency and portability.

calm. CUDA/Metal accelerated language model inference

luminal. Inference at the speed of light.

flashinfer. FlashInfer: Kernel Library for LLM Serving

UFO. UFO³: Weaving the Digital Agent Galaxy

GPTFast. Accelerate your Hugging Face Transformers 7.6-9x. Native to Hugging Face and PyTorch.

tensorrtllm_backend. The Triton TensorRT-LLM Backend

TransformerEngine. A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on Hopper, Ada and Blackwell GPUs, to provide better performance with lower memory utilization in both training and inference.

PALO. (WACV 2025 - Oral) Vision-language conversation in 10 languages including English, Chinese, French, Spanish, Russian, Japanese, Arabic, Hindi, Bengali and Urdu.

space. Unified storage framework for the entire machine learning lifecycle

gemma.cpp. lightweight, standalone C++ inference engine for Google's Gemma models.

LWM. Large World Model -- Modeling Text and Video with Millions Context

minbpe. Minimal, clean code for the Byte Pair Encoding (BPE) algorithm commonly used in LLM tokenization.

metavoice-src. Foundational model for human-like, expressive TTS

lectures. Material for gpu-mode lectures

gpt-crawler. Crawl a site to generate knowledge files to create your own custom GPT from a URL

mini-lsm. A course of building an LSM-Tree storage engine (database) in a week.

TaskingAI. The open source platform for AI-native application development.

c_std. Implementation of C++ standard libraries in C

GenerativeAIExamples. Generative AI reference workflows optimized for accelerated infrastructure and microservice architecture.

gpt-researcher. An autonomous agent that conducts deep research on any data using any LLM providers

TensorRT-LLM. TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way.

moondream. tiny vision language model

Depth-Anything. [CVPR 2024] Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data. Foundation Model for Monocular Depth Estimation

marker. Convert PDF to markdown + JSON quickly with high accuracy

MobileVLM. Strong and Open Vision Language Assistant for Mobile Devices

infinity. The AI-native database built for LLM applications, providing incredibly fast hybrid search of dense vector, sparse vector, tensor (multi-vector), and full-text.

chai. Modern embedded SQL database