I build computer-vision systems with graph-theoretic backbones and IoT products that work end to end.
Preparing for product-based SDE roles.
| LeetCode problems | Codeforces rating | CGPA · BCA, GLA University | Hackathon podium finishes |
|
Who What I build Focus areas |
Currently building 🔹 Java + DSA depth |
| Languages | Java · Python · SQL · JavaScript |
| AI / Computer Vision | PyTorch · OpenCV · NumPy · Pandas · scikit-image |
| Backend | FastAPI · REST APIs · Node.js |
| Databases | MongoDB · MySQL |
| Tools | Git · GitHub · NetworkX |
Occlusion-Robust Road Extraction & Graph-Theoretic Criticality Analysis for Urban Mobility
Pipeline
Satellite Image ──▶ Preprocessing ──▶ U-Net ──▶ Road Mask ──▶ Skeleton
│
Resilience ◀── Criticality ◀── Graph Healing ◀── Gap Detection ◀─┴─ Graph
| Stage | What happens |
|---|---|
| Input | Satellite imagery uploaded through a FastAPI service |
| Preprocessing | Image preprocessing and tensor normalization |
| Segmentation | Custom U-Net for pixel-level road masks, evaluated with Dice / IoU |
| Occlusion-aware training | Synthetic cloud, tree, shadow, and obstacle regions, including road-aware targeted occlusions |
| Skeletonization | Road mask reduced to a skeleton |
| Graph construction | 8-connected skeleton graph with endpoint, junction, and connected-component detection |
| Gap detection & healing | Distance and angular alignment, Union-Find (DSU), Kruskal / MST-based connections |
| Graph analysis | Shortest paths and betweenness centrality to identify critical nodes |
| Stress testing | Node-failure simulation, connectivity comparison, and alternate-route analysis, returned as JSON |
AI + IoT health monitoring: reports, vitals, and medication in one product
Medical Reports ─┐
Wearable Sensors ─┼──▶ HealthMate AI ──▶ Analysis · Monitoring · Assistance
User Interaction ─┘
| Reports & Insights OCR-based medical report analysis, symptom interpretation, preventive health insights |
Live Monitoring Heart rate, SpO2, temperature, and fall detection through wearable integration on a live dashboard |
Assistance Real-time AI chatbot using relevant historical user context, plus an emergency-contact workflow |
| Medicine adherence: scheduling and day-to-day tracking, with reminders on an I²C display and intake confirmed by a physical button | ||
IoT / embedded wearable for continuous monitoring and emergency response
A wearable safety prototype built around embedded sensors and a compact hardware interface, with sensor-driven event detection and automated emergency alerts.
| 🥇 | 1st Position · CodePunk V2.0 | HealthMate AI |
| 🥇 | 1st Prize · HackVision 2K26 | HealthMate AI |
| 🏅 | 2nd Runner-Up · Technavya Techfest | LifeGuard Band |
| 🏁 | Grand Finale Finalist · HackWithUttarPradesh 2025 | Hackathon |
| 🏆 | Top 25 · DEVIATHON National Hackathon | Hackathon |
| 🎨 | 1st Position · GitHub Logo Design Competition, DEVIATHON | Design |
-0d1117?style=for-the-badge&logo=microsoft&logoColor=58a6ff&labelColor=0d1117&color=58a6ff){kind=icon}
Building systems that keep working when the input is imperfect.