Cortex

Inspiration

Owning a home shouldn’t feel like guesswork, but for most people, it does. Critical information about your home is scattered across receipts, emails, and memory. When something breaks, you’re left reacting instead of planning.
One of our friends experienced this firsthand. An unexpected cold shower led to the discovery of a failing water heater, followed by days without hot water and a rushed replacement, but the warning signs were there. Things like age, service history, and expired warranty were on record somewhere in the house, but none of them were connected or actionable.
This gap inspired Cortex: An intelligent system that doesn’t just track your home, but understands it, anticipates issues, and helps you make informed decisions before problems escalate.

Cortex has 2 main pieces:

Knowledge Base: A structured database that maps your entire home: systems, devices, warranties, maintenance history, service providers, finances, etc. It is constantly updating to make sure that the ecosystem is as up-to-date as possible.
Predictive Models: Take in information from the knowledge base. Using this information it generates actionable data that is used all over the ecosystem including things like comparing repairing vs upgrading appliances, monitoring energy costs, etc.

Aggregated household data: systems, appliances, maintenance records, and expenses
Leveraged Gemini to analyze home market value using property data and external signals, as well as an intelligent advisor to answer ecosystem-level questions (maintenance, costs, and system health)
Built predictive models to detect anomalies and forecast failures (e.g., aging water heaters, rising energy usage)
Developed a reasoning layer that connects signals across our database to provide and justify explanations
Integrated service provider history to surface trusted contacts at the right moment
Created a feedback loop where user decisions improve future recommendations

Because Cortex is designed to be an ecosystem, it requires a massive amount of data to run efficiently. Unfortunately, that means that we had to handle a lot of data internally, which took time to process.

Our team worked together very well, and we were supported by some intelligent mentors. We started off this hackathon not knowing what to do, but once we received some direction, it was smooth sailing.

Building Cortex deepened our understanding of:

Designing AI systems that reason across fragmented, real-world data
Translating raw data (receipts, service logs, usage patterns) into actionable insights
Balancing predictive modeling with practical, user-facing decision support
Structuring home system data (HVAC, plumbing, appliances) into meaningful risk signals

Now that we have our base, we can start accruing users to build out our database to continue to train our predictive model. We can also work on expanding the reach of Cortex since we decided to start small for testing.