\\\Please register via this link to secure your spot\\\
Beyond vector search: Building reliable RAG for complex questions
with Pratiksha Parsewar, Senior Consultant, Thoughtworks and Teng Lei, Principal Data Scientist, Thoughtworks
Vector search is a useful starting point for RAG, but it breaks down when applications need to answer questions that span multiple documents, sources or relationships. At that point, simply tuning embeddings or increasing the number of retrieved chunks often isn’t enough. Engineers need to understand how different retrieval architectures address different failure modes—and how to know whether they actually improve results.
This session takes a practical, engineering-focused look at moving beyond naive RAG, covering approaches such as hybrid, multi-stage and graph-based retrieval, alongside a rigorous framework for evaluating them.
You’ll learn
Where basic vector RAG breaks down and how to identify its failure modes
How to choose between different retrieval architectures based on the problem you're solving
How multi-hop and graph-based retrieval can improve answers to complex questions
How to evaluate retrieval and end-to-end answer quality systematically
What to consider when moving more complex RAG pipelines into production
Who should attend: Software engineers, AI architects and data practitioners who are building RAG applications and want practical techniques for designing, comparing and evaluating retrieval architectures beyond basic vector search.
About the speaker
Pratiksha Parsewar is a Senior Application Developer and Agentic AI Engineer with 9+ years of experience across fintech, global banking, high-concurrency super apps, multi-agent AI architectures, and scalable microservices.
Teng Lei is a Principal Data Scientist and regional AI leader with 15+ years of experience across consulting, banking, AI engineering, analytics strategy, and enterprise transformation.
Agenda
6:30 pm Registration and networking
7:00 pm B