Drishti Jain

International Tech Speaker

BIO

Drishti is a Software Engineer and a technology enthusiast. She is currently based in the USA and believes in democratizing opportunities and brings knowledge of the latest developments in the fast-moving field of technology to early professionals, students, and keep them up to date and well-equipped for their professional careers. Her initiative “SkillUp with Drishti” enables students and professionals to grow in their career.

She has spoken at conferences across the globe and is also a social entrepreneur. Her non-profit organisation - Samyak Drishti Foundation works in environment, education and healthcare sectors and operates in a number of cities across India.

In her spare time, she likes to paint nature, explore new places and anchor live shows.

TITLE

Beyond the Vector DB: Building a RAG Stack That Actually Stays Fresh

ABSTRACT

Most Retrieval-Augmented Generation (RAG) systems start with a vector database—but that’s only part of the equation. In this talk, I walk through the real-world architecture I used to build and maintain an in-house RAG system that stays current with evolving internal documentation.

I’ll cover: Weekly refresh pipelines using checksum-based change detection to avoid redundant indexing Embedding versioning and drift tracking so model updates don’t silently degrade retrieval quality Semantic diffing and metadata tagging for fine-grained control over content updates Retraining workflows for embedding models and rankers that evolve with your knowledge base Freshness-aware retrieval, using timestamps and chunk-level versioning to prioritize recent content If you’ve deployed a RAG system but are struggling with stale results, hallucinations from outdated content, or slow content updates, this talk provides practical strategies to keep your AI knowledge pipeline sharp and reliable.