Chronel is a production RAG (Retrieval-Augmented Generation) platform that enables semantic querying over uploaded PDFs. Built a distributed 3-service architecture with independently deployed frontend, backend, and AI inference services on Render, enabling zero-downtime per-service updates and horizontally scalable async inter-service communication. Engineered the retrieval pipeline with chunk size 300 tokens, Pinecone top-k=5 vector search, and cross-encoder reranking to top-3 chunks before LLM context injection - improving answer relevance and minimising hallucinations. Measured end-to-end latency of 1.5-2s via LangFuse: 60% in vector retrieval, 40% in LLM generation.