AI Agents Conference · November 23–27, 2026
Sudhindra Desai
Session
From Prototype to Production: Architecting Agentic SDLC Platforms That Actually Ship
Most "agentic coding" demos stop at a clever prototype. Running one in production, inside a regulated enterprise with real compliance and on-call standards, is a different problem entirely. This session lays out the architecture patterns for taking agentic software delivery from demo to durable production system. It covers the full-lifecycle agent model spanning Plan, Design, Build, Verify, Secure, and Deliver, and the design decisions that determine whether such a system survives contact with real engineering teams. The talk goes deep on three pillars drawn from enterprise agentic AI research: a layered memory and context architecture that grounds agents in code history, regulatory rules, and team-specific procedures so their output is compliant and context-aware rather than generic; a model-routing and token-optimization strategy that keeps LLM costs from exploding as adoption scales; and a governance and traceability model that ties every agent action back to a business requirement, with the AI kept out of the critical path where determinism matters.
Share your speaker card
Speakers: download this card and post it on LinkedIn, X or Instagram. Links to this page preview with the same design.
Full conference agenda
See every session across the AI Agents Conference November 2026 line-up.