Scaling AI in the Enterprise - #SFTechWeek

Wednesday, Oct 8

3:00pm – 5:00pm

PTGMT

Hosted by

◕‿↼

ツ

ʘ‿ʘ

ᵔ▿ᵔ

•ᴥ•

:D

^◡^

+9

Sujith Joseph is a Principal Engineer and Chief Architect for Enterprise AI & Search in Cisco. He will be presenting about Cisco's Gen AI journey for drastically improving employee experience. Overview: Generative AI is everywhere; turning pilots into production at enterprise scale is the hard part. This talk goes behind the scenes of CIRCUIT—Cisco’s #1-used AI tool and enterprise agent ecosystem—now powering 100K+ employee interactions every business day and serving 4,000+ internal teams via LLM-as-a-Service and RAG-as-a-Service (powered by Pinecone Vector Database). We’ll unpack the search & cloud engineering that makes it work: an LLM router and agent orchestration layer (built with LangGraph) that routes tasks to the right expert agent/LLM; an AI Agentic Mesh with an Enterprise Agent/MCP Registry for discovery, governance, and reuse; and a security-first model using OAuth 2.0 Token Exchange (RFC 8693) so agents invoke one another with with secure identity propagation and a complete, auditable chain of custody. You’ll see how Deep Research Agents blend web + internal knowledge for actionable research reports; Personal & Cisco Data RAG enable safe BYO data (e.g., OneDrive); Data Analysis Agents accelerate insights; and a PPT Agent turns conversations into decks for faster storytelling. With Cisco AI Pods and Leading LLMs, employees can choose the best model for the job. We’ll close with outcomes from assistants in Sales (GSX), Cisco events (Cisco Live), support, and engineering—covering proposal generation, RFP compliance, and bug triage—and share the patterns to replicate this in your org. Takeaways • Reference architecture for agentic, multi-LLM platforms at enterprise scale • Patterns for reliable, stateful agents using LangGraph • Security model with OAuth 2.0 Token Exchange, agent identities, and audit trails • Operating model: registry, governance, and cross-agent reuse • RAG at Scale with Pinecone Vector Database • Practical playbook to move from POC to production with measurable productivity gains

•‿•

ツ

❛o❛