Design data infrastructure for AI agents & LLMs
Overview
Design AI agent-ready data platforms with safe write paths, state management, provenance, tracing, and production architecture.
Most AI courses teach you how to build an agent. This workshop teaches you how to build the data infrastructure that keeps one reliable in production. Through a single evolving case study, you'll design, break, and improve an AI agent platform while learning how to handle write operations, state management, provenance, recovery, and architectural trade-offs. By the end, you'll know how to evaluate whether a platform is truly ready for production AI—not just whether a demo works.
What You Will Get
Certificate of completion
Full HD recording of the live workshop
Downloadable PowerPoint presentation deck
Agent-Readiness Audit framework with four dimensions, twelve questions, and clear pass/fail criteria
Production failure-mode catalogue covering common infrastructure failures and their underlying mechanisms
Architecture review framework, design templates, provenance blueprint, and production checklists
Failure-mode catalogue covering real production issues and practical mitigation strategies
Why This Live Session, and Why Now
Learn through a continuous, hands-on case study instead of isolated coding exercises.
Explore the production challenges that appear after an AI agent works, including write conflicts, stale state, recovery, and observability.
Practice making architectural decisions, testing them in live labs, and defending your design choices.
Receive reusable architecture checklists and review frameworks that can be applied immediately within your own organization.
Understand infrastructure principles that remain valuable regardless of which AI framework or orchestration tool becomes popular next.
Tools and Frameworks you will learn
AI agent orchestration patterns
Python
Vector stores & embedding models
State management & memory design
Write-path design, retries & concurrency
Provenance, tracing & observability
Prerequisites
Comfortable reading and modifying Python code.
Experience with RAG or retrieval pipelines.
Familiarity with chunking, embeddings, and vector databases.
Basic understanding of schemas, pipelines, warehouses, and data lineage.
Some exposure to AI agents or multi-step LLM workflows.
This is an intermediate-to-advanced workshop and is not intended for beginners.
About the Speaker
Sandipan Bhaumik is a Data & AI Technical Lead at Databricks with more than 18 years of experience across data engineering, governance, analytics, machine learning, and AI. Formerly at AWS, he architects enterprise AI foundations and founded AgentBuild, a community focused on building practical, reliable, production-ready agentic systems at scale.