🧠 NeuroGuard

✨ Inspiration

NeuroGuard was born from a simple but urgent question: What if we could protect memory before it fades? 💭
I’ve witnessed the emotional toll of cognitive decline in underserved communities. I envisioned a system that could detect early stress markers in REM sleep—before symptoms appear—and do so ethically, accessibly, and scientifically. 🌱💡

🚀 What It Does

NeuroGuard is a predictive brain health platform 🖥️🧬 that analyzes REM sleep fingerprints using EEG and MEG data.
It identifies hippocampal stress patterns 🧩 and outputs a personalized cognitive risk score 📊.
The system is designed to run offline ⚡, on low-cost hardware 💻, with full explainability via SHAP 🛠️ and LRP 🔍.

🛠️ How I Built It

Modeled REM sleep features using public datasets from OpenNeuro 🗃️ and PhysioNet 📈
Trained a CNN-Transformer pipeline 🤖 to detect hippocampal stress markers
Built the backend with Python 🐍, TensorFlow ⚡, FastAPI 🚀, and PostgreSQL 🗄️
Deployed on AWS Cloud ☁️, with edge compatibility for offline-first use
Designed the frontend using HTML 🎨, featuring pastel themes 🌸 and glassmorphism 🪞 for emotional sanctuary

⚔️ Challenges I Ran Into

Public data was heterogeneous and noisy, especially EEG signals 📉
Balancing correlation vs. causality in stress markers like HRV and sleep spindles ⚖️
Ensuring explainability in a black-box model using SHAP 🕵️‍♀️ and LRP 🔎
Navigating ethical concerns around predictive health ⚕️ without definitive treatments
Advocating for visibility and fairness 🌍 in global competitions as a solo founder 🏅

🏆 Accomplishments That I’m Proud Of

Built a scientifically grounded model 🔬 with zero funding 💰 and full transparency ✨
Advocated for equity and visibility 🌏 in international platforms
Created a system architecture 🏗️ that prioritizes dignity, not just data
Delivered a 3-minute pitch ⏱️ that challenged the status quo of neurotech 💡