Hackathon Portal
Virtual Events
Team

aiio LeitWerk

Project Concept

By day, we build the process ontologies that agents need to stop guessing and start working — extracted from the real, unstructured chaos of how enterprises actually operate. This weekend, we get to push past it.
We’re a team from aiio GmbH in Magdeburg, Germany. Our daily work lives at the intersection of human, ai agents and real industrial processes — turning unstructured enterprise reality into structured ground truth that humans AND agents can actually act on.


The challenge that brought us here
Building the ontology is one thing. Letting humans navigate it is another problem entirely.
When you give people a system that captures process knowledge at real enterprise scale — multi-stakeholder, multi-source, versioned, governed — the interface becomes the bottleneck. Tree views collapse under the weight. Graph views look impressive and explain nothing. Chat interfaces hide structure behind a prompt. None of the conventional UI patterns were built for this: humans and agents collaborating on a living, structured representation of how an organization actually works.
That’s the wall we kept hitting. And it’s exactly where AG-UI and A2UI come in.


What we’re working on
We see AG-UI and A2UI as the missing layer — the first serious attempt at a protocol where agent state and human interface are designed for each other rather than bolted together after the fact. Our goal for the hackathon is to use that layer to prototype a new kind of interface: one where users can steer agents through complex, structured knowledge without drowning in it, and where agents can surface, propose, and execute changes that humans can review at the right level of abstraction.
Less “chatbot over a graph.” More control surface for human-agent processes.


How others can contribute
We’re looking for collaborators who are excited about:

  • AG-UI / A2UI protocol depth — anyone who’s been in the spec and wants to push its edges
  • Frontend craft — React / TypeScript folks who care about information density and interaction design, not just shipping another dashboard
  • Agent orchestration — multi-agent patterns, streaming state, human-in-the-loop flows
  • Knowledge representation & UX — graph UX, ontology navigation, structured editing
  • Skeptics — people who’ll tell us when our “next-gen UI” is just an old pattern in a new font
    If any of that sounds like your kind of weekend, come say hi.
    Autonomy needs a control surface. We brought ours — and we’re here to push it further.

* LeitWerk (German) — literally “guidance works” or “control assembly.” In aviation, the Leitwerk is the tail assembly of an aircraft: the rudder, elevators, and stabilizers that keep the plane on course. It doesn’t generate thrust — the engines do that. It steers. That’s exactly the role we see for the next generation of UI in agentic systems: the agents provide the power, the interface provides the control surface that lets humans keep the whole thing on course.

Entry

Status: Submitted

Last saved: May 09 at 11:14 PM PDT

Team Roster

Message board not available for this team yet.

Jobst von Heintze Team Lead RSVP Approved

Head of AI Innovation / CMO at aiio GmbH
PM - Claude, Claude Code used for concept check and creating synthetic datasets
CMO and Head of AI Innovation at aiio (Magdeburg) — a B2B SaaS company building agentic process intelligence for the DACH enterprise market. I sit at the seam between go-to-market and engineering: leading a cross-functional marketing/product innovation org, running enterprise BD with names like hyssenkrupp and ABB, and shipping production code in our monorepo at night. I work fluently across German and English, strategy and implementation.
Agentic systems for the enterprise — especially the governance layer between LLM agents and regulated processes (DORA, NIS2, EU AI Act). Knowledge graphs as the substrate for AI-native BPM. Causal inference and temporal reasoning in business decisions. Looking to connect with: founders building agentic infrastructure, engineers working on multi-agent governance, and consultancies who see process documentation as the missing prerequisite to enterprise AI rather than an afterthought.
Building ProcessForge — an AI-native, multi-user platform that turns SOPs, contracts, recordings and regulations into BPMN/SPEM process models in minutes. Compiler-pattern Canonical Process Meta-Model with pluggable adapters for third party tools. In parallel: ProcessCast (German audio explainers from process descriptions) and OpenClaw, a Claude Opus-backed agent with hardened security architecture. Shipping daily via Claude Code.

Ayesha RSVP Approved

AI Software Developer at aiio Gmbh
Dev - Claude Code and Cursor as IDE, focus on backend and KG genration
Ayesha Fiza is a Werkstudent in AI-based software development at aiio GmbH. She studies Digital Engineering and Computer Science at Otto-von-Guericke University Magdeburg and Acharya, with 3 years of experience.

Meet Patel RSVP Approved

AI Software Dev at aiio GmbH
Dev - Claude Code and VS Code as IDE, focus on frontend
I work in full-stack software engineering with a focus on building applications that integrate LLMs and modern web technologies. My current interests include developing intelligent systems that combine backend AI capabilities with interactive front-end interfaces. I’m currently looking to expand my skills in sensor integration, robotics, and computer vision, and I’m excited to explore how these areas can connect with AI-driven software systems.
Robotics, deep learning, computer vision, embodied AI, physical AI, world models
I am developing a chatbot system for Business Process Management (BPM) software that integrates knowledge graphs, Retrieval-Augmented Generation (RAG), and Large Language Models (LLMs) to process and understand unstructured data. The system structures enterprise knowledge to enable intelligent information retrieval and context-aware responses. I also use Mermaid diagrams to visualize business processes, system architecture, and data flow. This approach helps build an explainable and scalable AI