DejaWho

Inspiration

Our identity is reflected in how we think, react, and adapt. In conditions like Alzheimer’s, these patterns change long before clear symptoms appear. We wanted to explore whether everyday behavior could help people better understand and protect their cognitive identity over time.

What it does

DejaWho is a cognitive game platform that builds a behavioural profile based on how you play. Through short games inspired by real cognitive assessments, the app tracks patterns like reaction time, attention, and learning style. A machine learning model compares these behaviours to typical patterns and to your own history, looking for unusual changes. Based on this, the experience adapts, adjusting difficulty, pacing, and practice suggestions to match your cognitive identity as it evolves.

It’s not a diagnosis tool, but a way to understand and strengthen how you think over time.

How we built it

We built the frontend with HTML, CSS, Tailwind, and JavaScript, and the backend with Python Flask. We used scikit-learn for behavioral modeling and anomaly detection, and Ollama to power an in-app AI assistant.

Challenges we ran into

Turning cognitive science into fun, meaningful games Finding data that connects behavior to cognitive health Designing a system that respects identity without labeling or diagnosing usersexisting data

Accomplishments that we're proud of

Creating a personalized experience that adapts to each user Connecting identity, behaviour, and AI in a thoughtful way Building something with real-world relevance beyond the hackathon

What we learned

How identity can be modeled through behavior over time Using machine learning to adapt experiences, not just score users Building full-stack applications in a sensitive health domain

What's next for DejaWho

Long-term user tracking with a persistent database Stronger personalization models Validating our approach with more data and research

Built With

• css
• html
• javascript
• ollama
• python
• sklearn
• tailwind