Fitness App

Architected a compound AI backend combining Agentic RAG, LLM generation, and deterministic logic into three REST API endpoints using FastAPI and Pydantic DTOs.• Implemented a LlamaIndex RouterQueryEngine that dynamically routes queries to specialized vector indices using LLMSingleSelector for tool-based decision making.• Built an LLM-powered meal plan generator that calculates TDEE and macros in pure Python then injects targets into a structured Groq prompt to return personalized nutrition plans with exact gram amounts.• Developed a training program generator mapping user availability and goals to intelligent weekly splits (Push/Pull/Legs, Upper/Lower, Full Body) with strict JSON output validation.