Neysa CXO Unplugged

Most teams in this room have already wired LLMs into products. The frontier now is agents: autonomous workflows that talk to your APIs, orchestrate tools, and touch live customer and financial data. The question is not “can we build agents,” but “on whose infrastructure do they run, with whose weights, and under whose economic and security constraints. In this roundtable, we’ll go under the glossy demos and talk about the real engineering choices behind agentic systems in production. We’ll look at how Bangalore’s AI‑first startups are combining open‑source and open-weight models with proprietary components, mixing small, right‑sized models with heavier LLMs, and deploying them within sovereign environments rather than generic shared clouds. We’ll use Neysa’s Private AI Cloud and Velocis AI stack as live examples of how to run multi‑agent, multi‑model systems with dedicated GPU clusters, full‑stack control, and air‑gapped security, so your agents stay inside your perimeter while still moving at startup speed.