AI Research SKILLs
by Orchestra Research
researchadvanced
ai-researchmachine-learningmodel-trainingllmresearch-automationclaude-skills
A comprehensive open-source library of 86 production-ready skills covering the full AI research and engineering lifecycle — from literature review and experiment design through model training, optimization, inference, and deployment. Built by Orchestra Research and compatible with Claude Code, Codex, Gemini CLI, Cursor, and any agent using the SKILL.md standard.
Installation
npx @orchestra-research/ai-research-skills
Key Features
- ✓86 skills across 22 categories covering the full AI research lifecycle
- ✓Research orchestration: Autoresearch, Ideation, and ML Paper Writing skills
- ✓Training skills: fine-tuning, distributed training (Megatron-LM, DeepSpeed, ZeRO), post-training
- ✓Inference optimization: vLLM, TensorRT-LLM (24k tok/s), quantization
- ✓Safety and alignment: RLHF, constitutional AI, red-teaming, evaluation frameworks
- ✓Agent and RAG skills: multi-agent orchestration, vector retrieval, structured outputs
Use Cases
- →Autonomous AI research: literature survey → experiment → paper writing in one agent session
- →ML model development with best-practice training recipes (SFT, DPO, GRPO)
- →High-throughput inference optimization with vLLM and TensorRT-LLM
- →Safety-aligned model training with RLHF and constitutional AI pipelines
- →Production MLOps: monitoring, evaluation, and deployment automation