Build Night: The Data Layer for Agentic Apps
From demo-mode agents to production-grade systems.
Agents are moving from demos into production backends. That means they are no longer just generating text — they are reading, writing, remembering, and acting on real data.
Build Night is a new series on Agentic Engineering: building production systems where AI agents can safely read, write, and act on real data.
For Build Night #2, we are focused on the data layer for agentic apps. We are building on the pgEdge Agentic AI Toolkit for Postgres — enterprise distributed Postgres, agent-native database access via the pgEdge MCP Server, hybrid search with VectorChord-bm25 and pgvector, a turnkey RAG Server, and natural language querying. The full stack for moving agents from demos into trusted production.
We are gathering 50–80 curated engineers, founders, architects, and CTOs to work on the same primitives in the same room and see what people can ship in a few hours.
🛠️ What you'll do
15-min architecture walkthroughpgEdge engineering walks through the pgEdge Agentic AI Toolkit for Postgres: distributed Postgres for AI workloads, the MCP Server for agent-native database access, hybrid search and RAG primitives, and how it all fits together for production agent backends.
60-min build block Pair up or build solo. Clone the reference repo, extend it, break it, adapt it, or bring your own agent-touching backend problem.
Short lightning demos from the room, good conversation, and just hanging out and building together!
🎁 What you'll get
-
A working reference repo to clone and run during the night
-
Direct time with the engineers behind pgEdge and Constructive
-
A room of serious builders working on agent-touching backends and other cool stuff
-
Your build featured in the post-event reference publication if you demo. We have a video team on-site and will share all demo presentation videos with you after!
🧪 Suggested build prompts
Pick one or bring your own:
MCP-native agent access to Postgres - Build an agent that uses the pgEdge MCP Server to read, reason over, and safely write to a Postgres schema. Governed access, structured introspection, no glue code.
Cross-region agent writes - Build an agent that reads from one pgEdge region and writes to another. Use multi-master active-active replication and conflict-free delta-apply to converge concurrent updates without coordination.
Hybrid RAG on Postgres - Build a RAG pipeline combining BM25 ranking with pgvector semantic search on your own data, served via the pgEdge RAG Server.
Natural-language data interface - Build an agent that translates natural language queries to SQL using pgEdge's natural language layer, with audit and observability for every read and write.
Wild Card - Build something cool, build something random, and then show it off!
🕒 Format
5:30 PM — Doors open. Lobster rolls 🦞, drinks, and good conversation.6:10 PM — Welcome and opening remarks6:20 PM — Architecture walkthrough with pgEdge engineering 6:30 PM — Build block 7:30 PM — Lightning demos 8:00 PM — Network & chill
👥 Who this is for
Senior engineers, founding engineers, architects, and CTOs building or scaling agent-touching backends. This is especially relevant if you're:
-
Building a multi-tenant SaaS where AI features need per-tenant isolation
-
Running Postgres in production and hitting multi-region, HA, or data residency constraints
-
Designing the trust layer for agents that read and write production data
-
Evaluating distributed or agent-aware Postgres infrastructure for what comes next
Postgres is core to your stack, or about to be. You should be building, evaluating, or actively planning systems where agents touch production data.
Builders and hackers welcome!
🤝 About the hosts
pgEdge is enterprise distributed PostgreSQL for applications that need to stay highly available, low-latency, and consistent across regions. Its multi-master active-active architecture uses conflict-free delta-apply on numeric and aggregate columns — a CRDT-style approach that allows concurrent updates to converge without overwriting each other. pgEdge also builds the pgEdge Agentic AI Toolkit for Postgres, including the MCP Server, hybrid search via VectorChord-bm25 and pgvector, a turnkey RAG Server, and natural language querying. Customers include Bertelsmann, QRT, Jobot, the European Parliament, and U.S. government agencies.
Constructive builds secure-by-default Postgres for AI-native applications — multi-tenant agent-memory primitives that enforce isolation at the schema layer, so agents can safely read, write, and remember across customer environments without collapsing trust boundaries. The control layer between AI systems and production data — making tenant isolation, memory, permissions, and auditability part of the database architecture rather than application-level afterthoughts.
Frontier Syndicate is a venture community connecting frontier tech researchers, builders, and investors through curated convenings and early-stage capital. Across the Bay Area, we host a recurring series of research forums, builder nights, and intimate investor dinners — and back exceptional companies emerging from the labs, communities, and technical networks we convene.
AngelList builds the software investors use to fund world-changing startups. Founders Cafe is our founder community and co-working space on the first floor of HQ.