Every startup founder in the room has heard the "what" and "why" of agentic AI. This session answers the question that actually keeps them up at night: how do you build it so it doesn't fall apart at scale?

Kavya Hanisha spent years as a Senior Engineer inside Dell Technologies' AI division before becoming an AI Infrastructure Architect and Founder. She brings the infrastructure layer most agentic AI conversations skip entirely: the deployment bottlenecks that stall production, low-latency inference containerization via NVIDIA NIM microservices, and enterprise safety and policy enforcement using NVIDIA NeMo Guardrails. She’ll break down the real-world RAG pipeline decisions and MLOps architectures running across Dell and NVIDIA stacks that separate a fragile demo from a resilient, production-grade product customers can rely on.

Kavya holds a patent on automated systems analysis and brings seven years of core AI/ML systems experience to the stage. She is a Stanford WiDS 2026 Ambassador, an HBS Entrepreneurship alumna, and an active mentor to founders across Boston's startup community.

If your team is past the prototype and staring down the scaling problem, this session is built for that exact moment.

Key Takeaways:

• Production Multi-Agent Architecture: Patterns for orchestrating deterministic and non-deterministic agent workflows.
• High-Throughput Serving via NVIDIA NIM: Accelerating inference latency and scaling microservices across modern enterprise clusters.
• Enterprise Governance with NeMo Guardrails: Implementing programmable topical, safety, and execution boundaries for agentic tool use.
• From Lab to Datacenter: Overcoming real-world memory persistence, RAG latency, and observability bottlenecks in production.