Mindful Focus

Mindful Focus

Inspiration

Our inspiration is rooted in a critical and data-backed reality of the modern workplace: the dual crisis of employee burnout and Digital Eye Strain (DES). Reports from authorities like the Future Forum and Harvard Business Review are unequivocal burnout is an epidemic, costing the global economy billions and affecting over 41% of desk workers. Simultaneously, increased screen time has led to over 60% of computer users suffering from DES, directly harming focus and productivity. We saw a gap in the market. Existing tools are reactive and often require manual input. We envisioned a proactive, intelligent companion that doesn't just manage tasks but manages the user's well-being, creating a sustainable and healthy work environment.

What it does

Mindful Focus is a privacy-first, AI-powered desktop application that acts as your intelligent wellness companion. Using a standard webcam, it leverages 100% on-device machine learning to passively and securely monitor for early indicators of fatigue and eye strain. 1) AI-Powered Monitoring: It analyzes real-time metrics like your blink rate and relative pupil size to generate a "Focus Score" and an "Eye Strain" meter, providing a live look at your cognitive load. 2) Intelligent Break Suggestions: When the AI detects signs of fatigue, it sends a gentle, non-intrusive notification recommending a personalized micro-break, helping you recharge before burnout sets in. 3) Incentivized Wellness: At its core is our unique "Wellness-as-a-Service" (WaaS) model. Users subscribe to the service, but by following the app's recommendations to take healthy breaks, they earn points that translate into direct cashback on their subscription fee. The healthier your work habits, the less you pay. This transforms well-being from a chore into a rewarding experience.

How we built it

Our prototype was developed entirely in Python, selected for its robust AI ecosystem and rapid development capabilities, making it perfect for a 48-hour hackathon. 1) Core AI Engine: Our AI engine uses a sophisticated two-stage process for maximum efficiency and intelligence. a) On-Device Landmark Detection: We use Google's MediaPipe framework for real-time, on-device analysis. The OpenCV library captures video, and MediaPipe's Face Landmark detection processes it locally, tracking the precise coordinates of the user's eyes and face. This allows us to calculate high-frequency metrics like blink rate and head pose with incredible speed and privacy, as the raw video data never leaves the user's machine. b) VLM-Powered Contextual Analysis: The true intelligence is driven by Gemini 2.5 Pro, our Vision-Language Model. When our local MediaPipe logic detects a pattern suggesting fatigue (e.g., blink rate dropping while head pose droops), it securely sends the numerical landmark data and a single, relevant frame to the Gemini API. We then prompt the VLM to perform a holistic, contextual analysis, generating a personalized and empathetic recommendation (e.g., "I see you've been staring intently. Let's do a 20-second eye-focus exercise."). 2) Frontend User Interface: We built the desktop application's front end using Tkinter, Python's native GUI toolkit. Its simplicity and direct integration into our Python codebase allowed us to rapidly create a functional, lightweight interface to display the real-time "Focus Score," "Eye Strain" meter, and the personalized suggestions from Gemini. 3) Business Logic & Payments: Our innovative "Wellness-as-a-Service" model was brought to life by integrating the Unibee subscription management API directly into our Python application.

Challenges we ran into

The primary technical challenge was calibrating the AI model's sensitivity in a real-world environment. Factors like variable lighting conditions, reflections from glasses, and individual user differences made it difficult to get consistently accurate readings for pupil size. We overcame this by implementing a baseline calibration step on startup and focusing our logic on relative changes rather than absolute values.

Accomplishments that we're proud of

We are incredibly proud of building and demonstrating a stable, end-to-end functional prototype in just 48 hours. Every one of our acceptance criteria was met, from one-click webcam permission to a crash-free 2-minute demo.

What we learned

This project reinforced the power of a focused, lean development methodology. Our strict "Limit Work in Progress" rule was instrumental in our speed, forcing us to finish one component completely before moving to the next. Technically, we gained immense practical experience in the real-world application of on-device machine learning with TensorFlow.js and the challenges of building reliable computer vision features. On the business side, we learned how to rapidly prototype not just a product, but a unique monetization strategy, preparing us to pitch a compelling and validated case to potential investors.

What's next for Mindful Focus

This 48-hour prototype is just the beginning. Our vision for Mindful Focus is to become an essential tool for the modern workforce. 1) Enhance the AI Model: We plan to collect anonymized training data (with explicit user consent) to build a more robust, custom model that can detect a wider range of wellness indicators, such as posture and signs of stress. 2) B2B Enterprise Solution: We will develop a version for businesses, featuring an admin dashboard that provides anonymized, aggregated data on team well-being. This will allow companies to proactively address burnout at an organizational level while respecting individual privacy. 3) Deeper Integrations: We aim to integrate with calendar and productivity apps to provide even smarter recommendations, such as suggesting a short walk before a high-stakes meeting. 4) Expand the Reward Ecosystem: We will explore partnerships with wellness brands (gyms, meal services, meditation apps) to offer users more ways to redeem the points they earn for healthy habits.