using-llm-specialist

LLM task routing directs you to specialized skills based on your specific need—prompt engineering, fine-tuning, RAG, evaluation, context management, inference optimization, or safety. Describe your LLM task and the router's decision tree identifies the most relevant sub-skill and reference sheets.

Prompt engineering optimizes task performance by crafting input text, while fine-tuning adapts model weights using full training, LoRA, or QLoRA methods. Prompt engineering is faster and requires no data; fine-tuning requires labeled data but produces deeper model customization.

RAG system design integrates retrieval mechanisms with LLM generation to fetch relevant context and produce grounded responses. This skill routes you to RAG-specific guidance covering architecture, data handling, and integration patterns.

Safety and alignment tasks address model behavior guardrails, harmful output prevention, and value alignment during deployment. This router directs you to safety-focused strategies and evaluation methods for your specific risk profile.

LLM evaluation encompasses accuracy, relevance, coherence, and safety assessments tailored to your task. This skill routes you to evaluation frameworks and methods that match your deployment context and performance requirements.

Inference optimization reduces computational overhead through quantization, caching, batching, and context management techniques. The router directs you to optimization strategies suited to your hardware, model size, and latency requirements.