A comprehensive open-source library of AI research and engineering skills designed to supercharge AI models like Claude Code, Codex, or Gemini. This skill collection transforms your AI agent into a full-fledged AI research agent with complete horsepower.
Key Features
- 70+ Skills across 18 categories covering the entire AI research lifecycle
- 115,000+ lines of comprehensive documentation
- 58 gold-standard skills with extensive references and examples
- Full coverage from model architecture to production deployment
Skill Categories
- Model Architecture - Megatron-Core, LitGPT, Mamba
- Tokenization - Advanced tokenization techniques
- Fine-Tuning - Axolotl, PEFT, Unsloth
- Mechanistic Interpretability - Understanding model internals
- Data Processing - Efficient data pipelines
- Post-Training - Alignment and safety techniques
- Safety Alignment - Responsible AI practices
- Distributed Training - DeepSpeed, FSDP, Ray Train
- Infrastructure - Modal, SkyPilot, Lambda Labs
- Optimization - Flash Attention, GPTQ, AWQ
- Evaluation - Model assessment frameworks
- Inference Serving - vLLM, TensorRT-LLM, SGLang
- MLOps - Production ML workflows
- Agents - LangChain, LlamaIndex, CrewAI
- RAG - Chroma, FAISS, Pinecone, Qdrant
- Prompt Engineering - Advanced prompting strategies
- Observability - Model monitoring and tracking
- Multimodal - CLIP, Whisper, LLaVA, SAM, BLIP-2, AudioCraft
- Emerging Techniques - MoE, Model Merging, Long Context
Use Cases
- Transform Claude Code into an AI research assistant
- Build comprehensive AI agents with full research capabilities
- Accelerate AI research and development workflows
- Educational resource for learning AI engineering practices
Maintenance
Actively maintained by Orchestra Research with regular updates and community contributions. Licensed under MIT for open use and modification.
Quick Start
Simply integrate these skills into your AI agent's skill set, and it will gain comprehensive knowledge across the entire AI research lifecycle - from model architecture design to production deployment.