Late last year, I presented a talk on how to build a simple RAG assistant for your docs. Since then, I've moved beyond it to understand more about retrieval and enhancing the application. This talk focuses on retrieval engineering, advanced search and retrieval capabilities combining BM25, Vector search, Knowledge graph traversal with a RAGAS evals. I'll walk through some architectural and design decisions that moved the needle for my app: Luminary, and why phased ranking and cross-encoder re-ranking are where prod retrievals are heading in 2026.
Late last year, I presented a talk on how to build a simple RAG assistant for your docs. Since then, I've moved beyond it to understand more about retrieval and enhancing the application. This talk focuses on retrieval engineering, advanced search and retrieval capabilities combining BM25, Vector search, Knowledge graph traversal with a RAGAS evals. I'll walk through some architectural and design decisions that moved the needle for my app: Luminary, and why phased ranking and cross-encoder re-ranking are where prod retrievals are heading in 2026.
Senior Data Engineer
Anup Sethuram is a Senior Data Engineer at Foxtel Group/DAZN with 15+ years in enterprise software and data platforms. A DataEngBytes speaker and Brisbane co-organiser, he previously presented on streaming with Flink and building an AI assistant with RAG.
