# Compute, Latency, and Safety: Architecting Stateful Guardrails for Multi-Agent Workflows

PyTorch Conference North America 2026 · [All sessions](https://sessionize.com/api/v2/s2rq62nk/view/markdown)

- **Session ID:** 1251388
- **When:** Tuesday, October 20, 2026 11:45 am - 11:55 am (Pacific Standard Time, UTC-08)
- **Room:** LL21A (Lower Level)
- **Speakers:** Purva Chiniya

## Description

Deploying a single LLM is a solved infrastructure problem. Deploying a multi-agent system where autonomous nodes pass state, execute external tools, and orchestrate complex workflows introduces a massive architectural bottleneck: How do you secure inter-agent handoffs without destroying real-time throughput?

In production environments serving millions of users, standard security checks fail to catch "indirect orchestration exploits" (where untrusted data corrupts downstream agent privileges). But evaluating every single state change with a massive frontier model is computationally impossible and financially ruinous.

This session breaks down the MLOps infrastructure required to secure multi-agent architectures at enterprise scale. Drawing from the development of open-source evaluation frameworks like SecRepoBench, we will detail how to build continuous, stateful red-teaming pipelines. Attendees will learn how to deploy specialized, low-latency Small Language Models (SLMs) as deterministic orchestration guardrails, routing cross-agent payloads through high-speed evaluation checks that catch prompt injection and operational drift without bottlenecking the entire system.

## Speaker

- [Purva Chiniya](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f3b44840-a999-4c95-acef-bee81d34eeef) — Applied Scientist at Amazon Ads | Secure Agentic AI & Multi-Agent Systems