Looking back, written the day after the trip home from St. Gallen.
START Hack in St. Gallen: no prize, a good prototype
Straight from the road trip to START Hack: 36 hours on an AI-guided thesis journey for Studyond.
We didn’t win anything. Let’s get that out of the way.
On March 17 at 4 am, René and I got back to Berlin from the road trip. The same day I left for St. Gallen. With Ignite, the student founders’ initiative at HPI, twelve of us went to START Hack in three teams, with travel funding from HPI. START Hack calls itself Europe’s most entrepreneurial hackathon: more than 600 people, 36 hours, challenges from partner companies.
Our team was called Thesinator. The challenge came from Studyond, a Swiss startup: design and prototype an AI-guided journey through your thesis. With one clear rule. The AI suggests, the professor decides.
We built an entry flow with three questions, voice included, topic matching with embeddings on Supabase and pgvector, a multi-agent simulation that shows how a topic might play out, and a chat with your future self. My part was the chat, networking and browse features: more than 20 React components, two Supabase Edge Functions with their database migrations, and the persistence that lets the entry flow survive a reload.
One thing was new for me. Before we started, I condensed the challenge’s 89 context files into a compact knowledge base for our coding agents, one file they read first. The idea: the agents shouldn’t need the problem re-explained in every prompt.
The demo on this page shows how it fits together, with sample data. In the wizard you answer three questions, typed or spoken. Your answers become an embedding, a vector of numbers that describes what you’re after. pgvector finds the topics in the database whose vectors are closest, and they show up as cards you swipe left or right like in a dating app. After that the network view opens, with people, organisations and topics you can browse and ask questions about in a chat. Those network, browse and chat parts are what I built. Two Supabase Edge Functions answer the chat, and the session is saved, even if you reload in between.
Thesinator
Meet your future self through your strongest thesis paths.
Discover
Tell us what you want
Thesinator
Your thesis matches are ready
We found your best matches. Next, we show who is connected to them.
Showing your best matches while the full list loads.
Simulate your future
Building your stakeholder map
We’re connecting the people, thesis options, companies, universities, and experts linked to your best matches.
Current step Map ready
Latest update now
Stakeholder graph ready
Current step Future path ready
Latest update now
The live future path finished and is ready to explore.
Stakeholder map
See how people, organizations, and thesis options connect around your future path.
Selected connection
Click any point on the map to see who they are and why they matter for this path.
Explore theses
Choose a thesis to explore
Start with your best matches.
LLM Assistants for Maintenance Logs in Mid-Sized Manufacturing
Build and evaluate an assistant that turns free-text maintenance logs into structured fault reports.
Why this path now
It stays close to your interest in Technology & IT and keeps a company partner in the loop, which you said you want.
Could lead to: Applied AI Engineer at Northwind Robotics
Energy-Aware Scheduling for Edge AI Devices
Design a scheduler that trades model accuracy against battery life on small edge devices.
Why this path now
A technical path with a clear industry problem and room for your own prototype.
Could lead to: Embedded ML Engineer at Helio Grid
Validating B2B Ideas with Synthetic Customer Interviews
Test whether simulated interviews predict real customer answers for early B2B products.
Why this path now
It connects your startup ambition with a research question you can publish.
Could lead to: Founder in residence at the Venture Lab
Ask your future self
Ask what this path is like.
Why this path now: Swarm feedback from practitioners and supervisors says applied LLM work with a real partner…
You later on
From Maintenance Logs to Product: Your Five-Year ArcYou finish the thesis with a working assistant at Northwind Robotics, publish the evaluation and spend two years shipping it to more plants…
Your question
What does a typical week look like as an Applied AI Engineer at Northwind Robotics?
Future self
Most weeks start on the shop floor: you sit with two technicians, read last week's reports and see where the assistant got it wrong. Then you fix prompts and evaluation, not features. Fridays you demo to the plant manager. It is less research than your thesis and more about trust.
Try it: the demo runs by itself when it scrolls into view; click anywhere inside to take over. Sample data only.
Tools and skills: React and TypeScript with Tailwind and shadcn/ui, Supabase with Postgres, pgvector, Edge Functions and migrations, voice input and output in the browser, and preparing context for coding agents so they can actually help within 36 hours.
Why it wasn’t enough, I can’t say for sure. We certainly built a lot.
Now I’m back in Potsdam, after a bit more than three weeks away.