AI-Powered Hybrid RAG Platform for Telecom Optimization

Telecom engineering teams work with large amounts of operational and technical information, making accurate retrieval and grounded reasoning critical when using LLM-based assistants.

Designed and deployed a Retrieval-Augmented Generation system combining BM25 lexical retrieval with vector embeddings and cross-encoder re-ranking to improve retrieval quality and answer precision.

Implemented automated evaluation using faithfulness-oriented metrics, precision@k, and LLM-as-a-judge scoring to evaluate retrieval and generated responses.