If you thought ‘infinite context’ windows can effortlessly swallow your enterprise monolith, then think again. This session shatters the illusion of massive prompt windows and introduces a practical, distributed-systems approach to AI context management. Discover the exact framework used to orchestrate an army of specialised, tool-driven subagents that can successfully map out and reverse-engineer 10,000+ file XXL codebases without ever hitting the context wall.
Harshad will take you through how these specialised agents divide and coordinate the work, progressively build an understanding of a large codebase, and turn an otherwise overwhelming reverse-engineering problem into a structured, scalable process. The session will also explore the architectural choices that make this approach work in practice, and where the real challenges emerge when AI agents tackle enterprise-scale systems.
What you’ll learn:
How to manage AI context when working with massive enterprise codebases
How specialised, tool-driven subagents can divide and coordinate reverse-engineering work
A practical framework for mapping and understanding 10,000+ file XXL codebases
The architectural patterns that make multi-agent context management scalable
What it takes to move beyond a single agent trying to understand an entire codebase at once
Agenda
6:00 – 6:30 PM – Registrations
6:30 – 7:30 PM – Talk followed by Q&A
7:30 – 8:00 PM – Networking and Refreshments
