AI teams are burning tokens and budget faster than they're shipping value. This roundtable brings together engineering leads, ML platform owners, and AI product builders for a candid, off-the-record conversation on three interlinked challenges:

  1. Token Mileage, prompt compression, context window discipline, caching strategies, and cost-per-inference governance across production workloads.

  2. Harness Engineering, evaluation pipelines, regression testing for LLM outputs, and building internal tooling that lets teams iterate without breaking things.

  3. Guardrail Architecture for Agents, designing agentic systems with reliable constraints: tool-call limits, fallback logic, human-in-the-loop triggers, and output validation at scale.

Held under Chatham House Rule. No slides, no sales pitches, just practitioners comparing notes on what's working in production.

Organised by the Centre for Applied AI Governance (CAIG), in association with Bengaluru Tech Week 2026.