Building @ the intersection of deep learning research and high-performance production systems.
|
I bridge the gap between cutting-edge ML research and polished, production-grade full-stack systems — turning complex model architectures into real-world intelligent applications. My work spans seismic hazard intelligence, urban computer vision, industrial inspection automation, and climate-resilience AI. Name : Vansh Agrawal
Role : Full-Stack Developer & ML Researcher
Domain : Computer Vision · Urban AI · Structural Informatics
Location : Chennai, India (UTC +05:30)
Status : Building @ the intersection of research & deploymentCurrently researching a flood prediction model for the Bihar basin — applying hydrological ML to infrastructure resilience at regional scale. |
|
A global seismic hazard intelligence platform delivering real-time tectonic monitoring, probabilistic hazard scoring, and volumetric 3D tectonic simulations. Combines gradient-boosted ensemble models with interactive WebGL globe rendering. Architecture Highlights
|
An edge-ready visual intelligence framework that quantifies Urban Cleanliness Index (UCI) from municipal imagery. Combines real-time object detection with a generative augmentation pipeline for robust model training across diverse urban environments. Architecture Highlights
|
An enterprise-grade defect detection system developed during my Tata Steel internship for structural TMT steel bar manufacturing. Provides automated surface anomaly classification with sub-millimetre precision, reducing manual inspection overhead in high-throughput industrial pipelines. Architecture Highlights
|
An AI-driven emergency management platform that replaces reactive disaster response with proactive, ML-powered intelligence, built specifically for Bihar's annual flood crisis. Complements hydrological research with a live risk-tracking dashboard. Architecture Highlights
|
|
Multi-agent AI placement automation for B.Tech CSE students — autonomous job discovery, eligibility analysis, resume tailoring, and application tracking via a live dashboard. |
Real-time peer-to-peer WebRTC application for low-latency voice, video, and screen sharing — built for production-grade communication reliability. |
High-performance Retrieval-Augmented Generation platform. Upload PDFs and interrogate them via semantic search and LLM reasoning — minimal latency, maximum accuracy. |
|
AI-powered driver monitoring system that detects drowsiness and yawning in real time using computer vision and facial landmark detection. |
Full-stack mobile-first application designed for flatmates to fairly decide dinner using collaborative filtering and customized taste parameters. |
AI-driven smart parking space counter that processes overhead camera feeds in real time to classify parking occupancy and display available spaces. |
Core Languages
ML & Computer Vision
Full-Stack & APIs
Data, Infrastructure & Tools
|
🎵 Music is the other language I speak. Beyond systems and models, I maintain an active social presence dedicated to singing — a creative counterbalance to the precision of engineering. Music, like good architecture, is about finding the right structure for something that needs to breathe. |
📍 Chennai, India | UTC +05:30 Building systems that matter — from tectonic faults to urban streets to flood plains. Every line of code is an attempt to make the physical world a little more legible to machines, and a little safer for people. Open to research collaborations, internships, and interesting problems. |
"The most elegant systems are those that hide their complexity behind a clean surface."
