← All work

Hackathon ยท nwHacks 2025

Ctrl+Zen

Unlocking developer productivity with real-time attention and emotion tracking.

Role
Team lead, front-end, ML
Team
4 people
Timeline
24-hour hackathon sprint
Summary
Live facial-recognition productivity monitor
Stack
Python, OpenCV, DeepFace, TensorFlow, Flask, Java, Spring Boot, MySQL, HTML/CSS/JS, Figma

Ctrl+Zen on CBC News

An excerpt from the interview. Watch the full segment ↗

Inspiration

Have you ever banged your head against the keyboard for hours on end with nothing to show for it?

We're pretty sure every developer has been there; we certainly have. Ctrl+Zen draws inspiration from the simplicity of keyboard shortcuts and the universal need to stay focused when coding. We wanted to build a tool that promotes focus and productivity for developers, blending efficiency with mindfulness.

Overview

Ctrl+Zen is a facial-recognition productivity companion that manages your development workflow with data attuned to your emotional state. When we're agitated, we often push on instead of taking a breather. Ctrl+Zen is here to fix that.

By analyzing your expressions, the app detects when you are laser-focused or nearing burnout, then offers gentle nudges to take a break or power through. Over time, it compiles personalized insights so you can code more efficiently and accomplish more in less time.

How we built it

On the backend, we used Python for image analysis and Java for the database. OpenCV, DeepFace, and TensorFlow power the image analysis that turns a webcam feed into meaningful results.

On the front end, we designed the GUI in Figma and built it with JavaScript, HTML, and CSS, keeping the interface intuitive and distraction-free.

Challenges

Balancing functionality with simplicity was the biggest challenge. We had to prioritize features carefully to avoid overwhelming users while still delivering value.

Connecting data between the front end and backend was also difficult; after a lot of trial and error, we got it working with local servers through Flask. Configuring the computer vision models took similar trial and error to get accurate analyses.

Highlights

  • Got a full end-to-end pipeline working in a single weekend: capture, inference, and visualization.
  • Smoothed raw model outputs so the dashboard felt calm instead of jittery.
  • Balanced performance and readability so the UI stayed responsive on typical laptops.