Track 03 · Resilience
Assets That Speak Before They Fail
Making slow expertise instant, so sensors, roofs, suppliers and rooms speak before they fail.
Every challenge follows the same structure. Each track has a shared theme. Within it, each partner company has an arena: a real problem from their operations, a concrete build assignment, and a defined deliverable. Pick one arena and build for it. Whatever you build in one arena will be instructive to every partner in that track.
Background
Assets degrade quietly — a sensor, a roof, a hotel room, a supplier relationship. In every case the expertise to catch problems early exists, but it sits in processes too slow or too manual to act on: simulations that take hours, inspections that happen monthly, risk reviews that happen quarterly. The task in every arena is the same: make that insight instant, so the asset speaks before it fails.
Choose your arena
The problem
Manufacturing systems and renewable energy assets rely on sensors performing in complex physical environments. Design engineers use numerical simulations to evaluate how design choices and environmental conditions affect sensor performance — but those simulations are too slow for real-time, real-world use. The answer arrives after the moment it was needed.
Build
AI-accelerated physics models that predict sensor system behaviour instantly, showing how environmental conditions and design parameters affect sensor reliability, resilience, and performance in real-world applications.
Deliverable
An interactive dashboard powered by AI predictions, showing engineers and operators how environmental conditions and design parameters affect sensor reliability, resilience, and performance in real time.
Data sources
FT Technologies simulation run datasets.
The problem
Green roofs manage stormwater, reduce heat, and support biodiversity in Swedish cities — but once installed, they dry out, waterlog, or overheat long before a manual inspection catches it. The roof should be managed as a living climate asset, not a passive building feature, and today nothing makes that possible.
Build
A system that fuses roof sensor data, local weather forecasts and history, satellite imagery, and botanical reference knowledge to assess the health of a green roof in central Göteborg, flag risk signals such as drought stress, excess moisture, or heat stress, and recommend inspection or care.
Deliverable
An interactive "digital gardener" dashboard showing current roof condition, risk signals, and practical care recommendations for property owners, facility managers, or construction companies.
Data sources
NCC roof sensor data (relative humidity, temperature) from the site at 57.711546, 11.973253; open weather, satellite, and botanical reference data.
Dataset
The problem
Category managers sit on sourcing decisions worth millions, but the risk signals that should inform them — supplier financial health, geopolitical exposure, single-source dependency, contract exposure — live scattered across systems that don't talk to each other. By the time a disruption shows up in delivery data, it's too late to renegotiate.
Build
An agent that, given a supplier name, pulls together a risk assessment (financial health signals, geopolitical exposure, dependency concentration) and surfaces the contract obligations tied to that risk (termination clauses, SLAs, penalty triggers). Stretch: extend the agent to pull market pricing context or draft a negotiation brief — but a working risk-and-contract answer is the bar, not a full multi-agent system.
Deliverable
A chat interface where a sourcing manager asks "What's my exposure with Supplier X right now?" and gets a combined risk-and-contract-obligation summary in one defensible answer.
Data sources
A working set of open, real data sources mapped to financial health, geopolitical exposure, dependency concentration, and contract exposure.
Helpful resources
- SEC EDGAR — financial health (US-listed / ADR)
- OpenCorporates — financial health
- Alpha Vantage — financial health
- World Bank Worldwide Governance Indicators — geopolitical exposure
- OpenSanctions — geopolitical exposure
- UCDP (Uppsala Conflict Data Program) — geopolitical exposure
- GDELT — geopolitical exposure
- UN Comtrade — dependency concentration
- OEC (Observatory of Economic Complexity) — dependency concentration
- USGS Mineral Commodity Summaries — dependency concentration
- GLEIF — dependency concentration
- CUAD (Contract Understanding Atticus Dataset) — contract exposure
- SEC EDGAR full-text search — contract exposure
- World Bank Commodity Markets, Pink Sheet — market pricing (stretch)
- NY Fed Global Supply Chain Pressure Index — market pricing (stretch)
The problem
Heating and cooling unoccupied guest rooms is one of the largest line items in a hotel's operating cost and carbon footprint, but energy management systems apply one setback temperature to every room — ignoring sun exposure, orientation, floor level, shading, and season, all of which are knowable in advance. Two identical rooms can have completely different heating needs, and nothing in the system is listening to that difference.
Build
A system that uses publicly available environmental, weather, and geospatial data to recommend the optimal energy-saving thermostat setpoint per unoccupied room, predict recovery time back to guest comfort before check-in, explain each recommendation, and estimate energy savings versus a flat building-wide setpoint. Optimize at the room level based on floor plan and position — two rooms in the same hotel should be able to get different answers.
Deliverable
A working prototype that, given a room's location and current conditions, outputs a recommended setpoint, a comfort-recovery time estimate, a plain-language explanation, and an estimated savings figure.
Reference property
Ringvägen 98, Stockholm View on map
Assume floors 3 through 8 contain guest rooms and suites. Room layouts and positions should be derived from publicly available information and reasonable assumptions. Participants may create synthetic floor plans and room inventories based on the building geometry. The goal is not to model the entire property, it's to optimize thermostat settings at the individual room level.
Data sources
The challenge is built around publicly available and self-generated data. Participants may use these examples or find equivalent sources.
Weather & climate data: OpenWeather, Meteostat, NOAA, Copernicus Climate Data Store.
Solar & environmental data: solar position datasets, solar irradiation datasets, PVGIS, cloud cover and UV data.
Geospatial & building data: OpenStreetMap, public GIS datasets, public satellite imagery, building footprint and orientation data, nearby structures that may create shading.
Synthetic data created by participants: floor plans, room inventory, occupancy forecasts, check-in/check-out patterns, historical room temperatures, thermostat settings, guest comfort profiles.
Shared requirement for all teams
Your system must degrade gracefully when data sources are incomplete or low-resolution, and be explicit about prediction uncertainty.
Pitch
A short pitch covering the end user, where in the operation your solution deploys, and how it scales from one asset to many.