A curated collection of reusable skills, scripts, code examples, and pipelines for LLM engineering. From RAG to agents to inference optimization. Research-backed, production-tested.
Reusable prompts and techniques for LLM engineering
Utilities for data, model, evaluation, and deployment
Minimal to end-to-end implementations
End-to-end workflows for data, training, eval, deploy
Production-ready config profiles for models, runtime, and deployment
Comprehensive unit, integration, and end-to-end tests
Explore the codebase structure and discover resources
git clone https://github.com/Shuvam-Banerji-Seal/LLM-Whisperer.git
pip install -r requirements.txt
ls skills/*/
python sample_code/minimal/minimal_agent.py
pytest tests/ -v
Deep dive into each topic with curated resources, papers, and code
Retrieval-Augmented Generation: agentic RAG, graph RAG, multimodal, evaluation
LoRA variants, RLHF, DPO, GRPO, curriculum learning, model merging
Multi-agent systems, tool use, memory, planning, observability
Speculative decoding, PagedAttention, quantization, CUDA graphs, edge deployment
Chain-of-thought, tree-of-thought, graph-of-thought, MCTS, reasoning models
Alignment, red teaming, hallucination detection, PII redaction, guardrails
Recently added skills and resources
15 new skills: speculative decoding, PagedAttention, KV cache optimization, FP8/GPTQ/AWQ quantization, disaggregated serving, multi-LoRA serving, MoE inference, distributed inference, CUDA graphs, streaming generation, model caching, batch optimization, monitoring, edge deployment, cost routing
New skills: reasoning models (o1/R1-style), graph-of-thought, program-of-thought, multi-step reasoning, uncertainty quantification
New skills: self-RAG, RAG fusion, hierarchical RAG, RAG guardrails, advanced document parsing, chunking strategies, embedding optimization, real-time streaming RAG, agentic workflows, tool use patterns
New skills: AI alignment techniques, red teaming, hallucination detection, MLOps pipeline, load balancing, MoE training, long context models