Multi-Agent Systems at Scale as a Shared Platform for the enterprises

AI Agent Infrastructure as a Shared Platform: Patterns for Multi-Agent Systems at Scale for the enterprise.

LOCATION ADDRESS (Hybrid, in person or by zoom, you choose)
Valley Research Park
319 North Bernardo Avenue
Mountain View, CA CA 93043
Don't use the front door. When facing the front door, turn right along the front of the building. Turn left around the building corner. The 2nd door should be open and have a banner and event registration.

If you want to join remotely, you can submit questions via Zoom Q&A. The zoom link:
Zoom (updated 6:55 pm)
Join via YouTube:
https://www.youtube.com/watch?v=pO72Hb30fKw

AGENDA
6:30 Door opens, food and networking (we invite honor system contributions)
7:00 SFBayACM upcoming events, introduce the speaker
7:15 Speaker presents.
8:30 - 8:45 finish, depending on Q&A

Join SF Bay ACM Chapter for an insightful discussion on:

### Abstract & Overview

An agent is simple: Prompt + Tools + Model + Boilerplate. The first three are where product teams create value. The last one—state management, history compression, streaming, cancellation, tracing, memory, persistence—is 80% of the code but 0% of the differentiation.
At ThoughtSpot, we built an Agent Platform that draws a hard line between agent logic and agent infrastructure, letting product teams ship customer-facing agents faster by owning only what matters: their prompts and their tools.
This talk covers the infrastructure patterns behind that separation:
State management across tool calls. Stateless tools (state on the agent, passed as arguments) give you testability and let the LLM reason about state. Stateful tools (state in the tool service) avoid serialization overhead. I'll walk through flow diagrams, show how we propagate state via tool response metadata, and discuss when each pattern fits.
Configuration-driven agent definitions. Agents defined entirely through config—templated prompts, tool endpoints, sub-agent rules, compression strategies. Teams ship agents without writing orchestration code.
Inter-agent communication. Two patterns: agents-as-tools (sub-agent called like any tool, returns structured output) and agent handoff (full conversation transfer). The platform handles routing and context—teams just declare delegation rules.
Shared memory across agents. Memory in the platform, not individual agents, means knowledge accumulates across agent boundaries. Tiered scoping (tenant, org, user) with retrieval that surfaces relevant context regardless of which agent captured it.
Tool protocol design. MCP as the base, with patterns layered on top: cancellation semantics, progress streaming, context variable propagation, and adapters for existing services.
Building for customer-facing scale adds constraints—high concurrency, encryption, tenant isolation, auditability—that shaped our API design throughout.
Takeaways:

• Mental model for separating agent value from infrastructure
• State patterns: agent-side vs. tool-side tradeoffs
• Inter-agent communication: tools vs. handoff
• Shared memory architecture across agent boundaries
• MCP extensions for production systems.

Speaker Bio
Ashish Shubham is Fellow/Vice President of Engineering at ThoughtSpot, where he leads the architecture of enterprise-scale AI and embedded analytics platforms used by Fortune 500 organizations. He is the author of Architecting AI Data Systems and an inventor on multiple U.S. patents in natural-language-to-SQL, generative AI interfaces, and intelligent analytics. Ashish is an IEEE Senior Member and an active reviewer and committee contributor for leading IEEE and ACM conferences and workshops. His work bridges academic research and real-world deployment, with a focus on building scalable, trustworthy, and developer-centric AI systems for production environments.
https://linkedin.com/in/ashubham

---
Valley Research Park is a coworking research campus of 104,000 square feet hosting 60+ life science and technology companies. VRP has over 100 dry labs, wet labs, and high power labs sized from 125-15,000 square feet. VRP manages all of the traditional office elements: break rooms, conference rooms, outdoor dining spaces, and recreational spaces.

As a plug-and-play lab space, once companies have secured their next milestone and are ready to expand, VRP has 100+ labs ready to expand into.
https://www.valleyresearchpark.com/