Inspiration - In today's hyper-connected digital workspace, knowledge workers face an invisible epidemic: constant context-switching and mental fatigue. Every notification, tab switch, and background process drains working memory, leading to what cognitive scientists call attention residue and chronic burnout.
Sovereign Vault was born from a simple question: What if an AI could act as an external cognitive shield, monitoring mental load in real time, cutting through digital clutter, and delivering sharp, Socratic guidance to help professionals shed unnecessary stress before burnout hits? We wanted to build a tool that treats mental well-being with the same rigorous telemetry usually reserved for server infrastructure.
What it does - Sovereign Vault is an interactive, light-mode cognitive health monitor and load-shedding engine. It offers:
Live Telemetry & Stress Scoring: Tracks working memory strain, analyzes active friction sources, and visualizes mental load trends in real time.
The Socratic Guardian: Powered by a fine-tuned AI model, it executes real-time load-shedding protocols to provide tailored, practical interventions that restore focus.
Team Stress Dashboard: Evaluates burnout thresholds and stress scores across a simulated 100-persona professional cohort to provide aggregate team well-being insights.
Secure Telemetry Vault: Features an integrated feedback portal that logs user testing data securely into a structured Google Sheets backend.
How we built it - Intelligence Core & Search: We fine-tuned meta-llama/Llama-3.1-8B-Instruct via Nebius AI as the core reasoning engine for cognitive reframing, integrated with Tavily AI to dynamically fetch real-time productivity and cognitive science research to ground the Socratic Guardian's interventions.
Frontend & Design System: Developed using Streamlit with a custom CSS design system tailored for high-contrast light-mode aesthetics, clean card layouts, and responsive data streams.
Data & Simulation Backend: Built using Python and Pandas to manage local database storage and simulate professional cohort telemetry matrices.
Integration & Deployment: Deployed via Streamlit Cloud, utilizing a custom Google Apps Script Web App endpoint to securely handle feedback submissions.
Challenges we ran into - Streamlit UI Customization: Overriding Streamlit's rigid component styling and Shadow DOM to achieve a crisp, high-contrast light-mode aesthetic required targeted CSS refactoring.
Model Fine-Tuning Calibration: Tuning the Llama-3.1-8B-Instruct model to deliver sharp, actionable, and non-repetitive cognitive load-shedding advice demanded careful prompt design and parameter adjustments.
Serverless Data Pipeline: Configuring the Google Apps Script Web App to safely accept POST requests from the Streamlit frontend without triggering CORS or permission errors.
Accomplishments that we're proud of - Successfully fine-tuning and deploying an open-weight Llama model on Nebius AI into a production-ready, highly polished web application.
Transforming complex psychological and system telemetry into an intuitive, low-cognitive-friction dashboard that anyone can understand at a glance.
Establishing a fully functional, end-to-end feedback loop directly connected to Google Sheets.
What's next for Sovereign Vault - Automated OS Integration: Connecting directly to screen time and active window tracking to spot digital distractions in real time. Instead of intrusive control, the system uses ethical design to offer gentle, timely nudges—guiding individuals away from constant multitasking and toward healthier habits, better focus, and an overall balanced lifestyle.
Advanced Predictive Analytics: Scaling the team stress matrix to help organizations forecast burnout risks across multiple departments and take proactive steps early.
Log in or sign up for Devpost to join the conversation.