Date: December 8, 2026
Time: 1:00 PM – 7:00 PM
Location: AWS London Office
Overview
Most AI applications end up running on several databases: a vector store for embeddings, a relational database for application data, something else for conversation history. Stop maintaining all four. Watch a multi-store AI stack break, fold it into one unified data layer, and ship a governed agent in one afternoon.
Join AWS and TiDB at London AWS Holborn for an afternoon of building, pizza and beer. We will open with a short, interactive talk on data and the danger of hallucinations. Then, we go hands-on: First, deploy a database with real data while learning how vectors affect Agent performance. Build two applications and observe why single data planes are essential. You’ll set up Agent runtime on AgentCore and build an AI customer support bot on top of it with Strands Agents, Bedrock, evaluations, and policy guardrails on top. The engineers who build these stacks will be in the room with you.
Then we close the laptops for pizza, beer and conversation.
What you’ll build
Two hands-on labs, two working apps. Learn why vector similarity and agent harness are indispensable. Leave with working apps built on your own data & policy guardrails.
Lab 1: RAG on your own data, fold three databases into one.
Start with a bare model and watch it hallucinate. Add vector search, then full-text and hybrid search across MySQL, Milvus and Elasticsearch, and learn where a multi-store frankenstack breaks. Then, we run the same app on a single TiDB table, with vector, full-text and hybrid search in one unified layer. Deep dive into vector search and why similarity matters. Models run via Amazon Bedrock.
Lab 2: Build a customer support bot
Build an AI customer support bot on Amazon AgentCore. Scaffold an agent with Strands Agents and frontier models. Add custom tools, policies, runtime & conversational memory through AgentCore. Continue with authentication, evaluations and policy governance.
If you can write SQL and call an API, you can complete both labs.
Bring: Just you and your laptop. AWS accounts are provided.
Optional: a free TiDB Cloud Starter instance if you want your playground to persist after the event.
Why attend
Meet the builders. Close the day with pizza, beer and the engineers building AI on AWS in the city.
Leave with working apps and industry knowledge, not a slide deck. You’ll go from an empty setup to a running AI customer support bot in one afternoon.
Engineers from AWS and TiDB designed the workshop and will be in the room to help you through it.
See one data layer replace four. Vectors, transactions and conversation history in a single database, so there’s nothing to keep in sync.
Programme
13:00 – Welcome
13:15 – PingCAP and TiDB
13:45 – AWS: how it fits, better together
14:15 – Break
14:30 – Lab 1: TiDB Cloud
16:30 – Lab 2: AWS AgentCore
18:00 – Pizza and beer
19:00 – Close

Anja Pettersson Solutions Architect at AWS
Anja Pettersson is a Stockholm-based AWS Solutions Architect specialising in agentic AI. She helps Nordic software companies modernise their platforms and bring AI into their products.

Bernard Kavanagh Principal Solutions Architectm Applied AI, TiDB
Bernard is a Principal Solutions Architect focused on Applied AI, designing production architectures for agentic and AI-native workloads. He turns the TiDB + AWS joint value into reference patterns engineering teams can actually deploy — collapsing the transactional database, vector store, cache and analytics stack into one engine on AWS, and showing where it fits for payments, multi-tenant SaaS, real-time risk and agentic features.