Skip to content

Details

## Build an AI Backend

​Chatbots are frontend. The interesting question is: what happens when AI moves into your backend?

​Not a copilot. Not a chat widget. A reasoning layer that sits alongside your services - making decisions that used to be hardcoded, replacing brittle if-else chains with logic that actually understands context.

AgentField is infrastructure for this. Agents that share memory without Redis. Agents that find each other without service registries. Just Python/typescript functions - no DAGs, no YAML, no workflow definitions.
​Come build something that belongs in a backend, not a chat window.

The Workshop:
​In one hour, you'll build a system where multiple agents research a topic in parallel, share what they find, and go deeper where it matters.

​Not by wiring up message queues. Not by defining DAG configs. You'll write Python (or typescript) functions, and the coordination just happens.

  • Reasoners - Functions where AI makes decisions
  • Skills - Functions where code does the work
  • Memory - State that flows between agents automatically
  • Discovery - Agents call each other by name, routing happens

​By the end, you'll have a working deep research system - the same pattern that powers everything from due diligence pipelines to adaptive content generation.

​Want to try it before the hackathon? Join our Discord - ask questions, get help if you get stuck, or just say hi.

What makes this different:

  • Memory without infrastructure - Memory that works across all your agents. No Redis, no Postgres, no config.
  • Agents find each other - Discovery engine routes automatically. No service registry setup.
  • Reasoners vs Skills - AI decides, code executes. Clean separation that's easy to test and reason about.
  • Just simple - No workflow DSLs. No DAG definitions. Write functions, decorate them, done.

The Challenge:
​Rethink what AI applications look like after the chatbot.
​The chatbot was AI-as-interface: human inputs, AI responds. The AI backend is AI-as-infrastructure: agents that make decisions, share context, and do work - whether a human is watching or not.
​Build the application that exists on the other side of that shift.

🏆 Prizes:
​$2,000 USD in OpenRouter credits:

  • ​1st Place: $1,000
  • ​2nd Place: $600
  • ​3rd Place: $400

⚡ Agenda:

  • ​10:00 AM - Kickoff + Workshop: From Hello World to Deep Research
  • ​10:30 AM - Team formation + hacking begins
  • ​12:00 PM - Lunch
  • ​3:30 PM - Hacking ends / Live demos (3 min per team)
  • ​4:30 PM - Winners announced
  • ​5:00 PM - Close

Who should come:

  • ​You've built chatbots and want to build something deeper
  • ​You're curious what AI looks like when it's not a UI feature
  • ​You want to replace some hardcoded logic with reasoning
  • ​You just like building things

​Python or TypeScript. No agent experience needed.

How we'll judge:

  1. New problem space - Is this AI-as-infrastructure in a domain we haven't seen? Show us something beyond chatbots.
  2. Replaced complexity - Did you simplify an architecture that was previously a mess of services, queues, or hardcoded rules? Does it make sense cost-wise?
  3. High leverage - Either high volume (worth automating) or high stakes (worth reasoning about). Bonus if both.
  4. Previously impossible - Does this enable something that couldn't exist before? Not just faster or cheaper - genuinely new.

💡 What could you build?

  • Pricing that adapts in real-time - Considers inventory, demand, and competition (for example) simultaneously. Reasoning and autonomous, not rules.
  • Supply chain that coordinates - Shortage detected, alternatives evaluated, routes recalculated, alerts sent.
  • Code Security Review for PR - Every PR launching and finding complex security threats and patterns that are non-trivial

​Build something we haven't thought of here, and you have your own startup at hand !

📚 Resources:

🤝 Hosts:

  • AgentField - Backend infrastructure layer for autonomous AI.
  • AI Builders - Community of engineers building the future of AI.

⚠️ Limited spots. Register now.

Related topics

You may also like