The LLM Engineer Handbook catalogs the full lifecycle of large language model engineering, from pretraining to prompt management, guiding engineers beyond demos to production-ready LLM apps.
Nougat is Meta’s neural OCR system for academic PDFs, extracting LaTeX math and tables into structured Markdown using a Vision Transformer encoder-decoder. It offers CLI, API, and training tools.
OverlapNet uses Siamese networks on 2D range images from 3D LiDAR to detect loop closures by predicting overlap and relative yaw angle simultaneously. Practical demos included.
SafestClaw uses classical ML pipelines and local AI models to deliver 90% of OpenClaw’s features at zero cost, avoiding prompt injection and cloud dependencies.
vLLM Compressor applies advanced quantization and compression techniques to large language models, enabling optimized inference without requiring full model definitions.
YuE is an open-source Python foundation model for generating complete songs from lyrics using a two-stage architecture and audio in-context learning. It supports style cloning and LoRA finetuning under Apache 2.0.
Command Code uses a meta neuro-symbolic AI ’taste-1’ model to continuously learn and adapt to your coding style, enabling personalized full-stack project building and bug fixing.
GoMLX is a Go-native machine learning framework offering automatic differentiation, multi-backend support including OpenXLA acceleration, and ONNX compatibility. It enables training and inference of LLMs like GPT-2 entirely in Go, with a pure-Go backend for WASM.
4DGen extends Stable Video Diffusion to generate geometry-consistent multi-view RGB-D videos from single RGB-D inputs using pointmap latents. Trained on multi-view robotic datasets, it enables robot pose extraction from generated videos.
Action100M provides a hierarchical Tree-of-Captions annotation for 100M video segments, enabling multi-scale video understanding with LLM-generated captions. Explore its structure, tech strengths, and how to access the data.
Qwen3.6 from Alibaba uses gated delta networks and sparse Mixture-of-Experts to achieve near-397B parameter model performance with only 3B active parameters, supporting 201 languages and 262k context length.
There Is No Spoon offers a unique ML primer for software engineers, using physical analogies to build deep intuition for neural networks and architectures beyond memorization.
ChatTTS is an open-source conversational text-to-speech model trained on 100,000+ hours of bilingual audio. It offers fine-grained prosodic control and employs intentional quality degradation to prevent misuse.
A community-curated list of free machine learning courses on YouTube that organizes and vets educational content for practical learners and enthusiasts.
Explore the architecture behind X’s For You Timeline recommendation system, built on Scala, Rust, and advanced ML models. Understand candidate sourcing, neural ranking, and filtering pipelines.
A curated Jupyter notebook guide for machine learning interview prep at FAANG companies, covering coding, system design, and agentic AI systems added in 2025.
OVIE trains novel view synthesis models using unpaired internet images, avoiding the need for calibrated multi-view datasets. It uses Vision Transformers and foundation models for pose and depth encoding.
PokieTicker combines rule-based filtering, LLM sentiment analysis, and XGBoost prediction in a full-stack stock analysis app. Runs locally with no API keys.
Explore a 100-day machine learning coding challenge combining classical algorithms, deep learning, and curated resources. A practical, day-by-day learning path for self-directed devs.
NVIDIA Warp lets you write Python functions JIT-compiled into CUDA kernels for GPU-accelerated differentiable physics and ML integration, simplifying GPU programming in Python.
