How to build highly scalable vertical agents for enterprises

While current AI agents perform well on narrowly scoped tasks, they struggle to scale for enterprise applications. The primary challenge lies in orchestrating complex interactions between agents and their sub-agents/tools to solve long-horizon problems. This talk presents a framework for building autonomous agents capable of sustained, multi-hour agentic interactions that effectively tackle extended tasks from initiation to completion.
Who is this presentation for?
product managers, developers, any one who wants to learn how to build scalable AI solutions
Prerequisite knowledge:
Basic knowledge of LLM and how to build agents
What you'll learn?
How to build fault tolerant and scalable agents

Sujit Khanna leads at the intersection of AI and quantitative research in financial markets, building systems that redefine how markets synthesize knowledge. His work fuses long-horizon multi-agent systems, post-training reasoning models, and knowledge-graph architectures—creating frameworks that unite quantitative rigor with fundamental insight. A hands-on architect who codes alongside his high-performance team, he transforms vast data streams into actionable intelligence while deploying institutional-grade systems at startup speed—bringing cutting-edge AI research to production in weeks, not quarters.