Hackathon

Prizes

• 

  Read more

  You and your team will receive a 1.000 € Amazon Gift card (total amount for the team)

  Show criteria Now available

  Resilience Room
  An organisation of AI agents that decides under pressure, argues with itself and knows when to ask a human.
  24 hours. Three agents minimum + one critic agent. One decision you can actually defend.
  
  Real decisions are a negotiation, not a calculation
  When a city faces a heat wave, or a company a risky change, the answer is never a single calculation: it is a negotiation between whoever holds the data, whoever knows the constraints and whoever carries the responsibility.
  Pick the harder half of the problem: not the answer, but how the organisation of agents got there.
  
  What you have to build
  All six are mandatory, whichever scenario you choose

  1. Three distinct mandates
  2. A critic that can say no
  3. Escalation to a human
  4. Transparent agent dialogue
  5. External data by design
  6. Re-plan on a new event

• 

  Read more

  You and your team will receive a 1.000 € Amazon Gift card (total amount for the team)

  Show criteria Now available

  Logistics Grand Prix

  Step into the role of a European freight dispatcher: given real European cities, a mixed fleet of trucks and a list of transport orders, build the plan that tells every truck which orders to pick up and deliver, and in what sequence.

  Respect truck capacities and special cargo requirements while combining shipments cleverly to deliver as many orders as possible with the lowest cost, the fewest empty kilometres and the smallest CO₂ footprint.

  Submit your plan through our API, get an instant score out of 1000, and compete for the top of the leaderboard.

• 

  Read more

  You and your team will receive a 1.000 € monni card (total amount for the team)

  Show criteria Now available

  KON Dash Control Room

  KON Dash is Konverto Lab's enterprise LLM platform. We'll provide you with a realistic dataset showing how our customers use the platform and give you 24 hours to design and build a dashboard that our team can use to monitor them. If you want to take it a step further, you can also rethink the KON Dash user experience. The best solution will become part of our product roadmap.

• 

  Read more

  You and your team will receive a 1.000 € Amazon Gift card (total amount for the team)

  Show criteria Now available

  Calibrating Building Energy Models
  By improving building energy models to better reflect reality using energy consumption data, we can better predict energy and CO2 savings through retrofits.
  
  Challenge
  Find the best optimization algorithm
  Teams will find the best optimization algorithm (e.g. Particle Swarm Optimization, Bayesian Calibration, Genetic Algorithm) and associated hyperparameters to minimize error between simulated and actual energy use.
  
  Teams will be given the following:

  • A Jupyter notebook to easily run energy simulations using our Python library, PyBuildingEnergy.
  • Consumption data for 5-10 buildings.
  • Building geometry, building parameters, and weather files for target buildings will be pre-programmed.

  Challenge scoring

  • Teams will submit their best calibration algorithm and set of hyperparameters.
  • The key metric will be percentage difference in real versus simulated energy consumption, with monthly resolution.
  • Teams accrue points for each percentage difference in real versus simulated consumption, for each of the buildings in the challenge.
  • The team closest to zero wins – this shows the closest calibration to the real energy consumption of the buildings

• 

  Read more

  You and your team will receive a 1.000 € Amazon Gift card (total amount for the team)

  Show criteria Now available

  Every robot comes with a lot of documentation, and the answers people need are buried somewhere inside it. Your mission is to dig them out.
  
  You’ll start from a curated collection of Markdown files and images (with transcriptions) covering a range of robots and accessories: technical documentation, product sheets, specifications, functional features, and usage details. Distilled from publicly available PDFs, this collection is your condensed source of truth.
  
  Your challenge:
  turn this documentation into a system anyone can talk to. A technician hunting for a specific parameter value and a salesperson trying to figure out which robot fits a customer’s needs should both walk away with an answer that is accurate, clear, and grounded in the data.
  
  RAG (Retrieval-Augmented Generation) and LLM-generated wikis are today’s go to approaches, and a good place to start, but the architecture is your call. Pick it, build it, and defend it.
  
  We’ll judge it on three things:
  the quality of the answers, how easy it is to use, and how faithfully it sticks to the source.
  
  Whatever you build, it has to reach real people: your solution must include at least one working interface that end users can use to ask questions and get answers.
  
  Bonus points for going further:
  Plug-and-play retrieval. Keep your retrieval engine separate from the interface and expose it so other frontends can use it too: a web app, a chatbot, a voice assistant, or an existing tool.

  Speak every language. Let users ask in their own language, even when the source is written in another, and get a coherent answer back.