3 IEEE Systems.Edge-to-Cloud.
I build inference systems for resource-constrained environments — ML pipelines that run at the edge, IoT sensor networks that operate offline, and full-stack products that ship to real users.
My work spans three domains: AI/ML (EmotiScan, Agri-Guard — 2 IEEE papers), IoT & embedded systems (AgroFog+, Emergency Ventilator — 1 IEEE paper), and full-stack engineering (LifeLineX, this portfolio).
4 hackathon wins — Harvard · Stanford · SRM Delhi · HackUEM 2.0
Available: Research / Applied ML Internship · Summer 2026 · Remote or Bangalore / Delhi
Building Systems That Actually Ship
10 projects across 6 domains. Not just models — deployed applications with live URLs, real users, and measurable impact.
Customer Churn Prediction
Ensemble model (XGBoost + LightGBM + CatBoost voting classifier) on IBM Telecom dataset. Deployed on HuggingFace Spaces via Gradio. Trained with SMOTE oversampling to handle 14% churn class imbalance.
Smart Retail AI
AI Trend Analyzer + Dynamic Price Optimizer. Uses scikit-learn regression for price elasticity modeling. Inputs: competitor pricing, inventory levels, seasonal index. Output: margin-optimized price recommendation.
FoodLens
REST API: POST /analyze accepts food image, returns macro breakdown. CV model identifies food item via fine-tuned ResNet, LLM layer generates nutritional estimate. Flask backend, 3-second average response.
Vedic Asharam
Service Worker offline cache (cache-first for content, network-first for video). 42 lessons, 6 modules. Web Speech API for audio. FCP <2s on 3G. Zero backend — pure JAMstack, $0 hosting cost, screen-reader accessible.
Military Support System
System architecture research: ESP32 wearable → MQTT broker → geospatial dashboard. Thermal CV module for anomaly detection. Designed for offline-capable field deployment at squad scale.
CustomCouture
Personalized fashion recommendation engine built for Myntra Hacker Ramp. Collaborative filtering + content-based hybrid model. Style profile generated from 8 preference signals. Targets 72% recommendation acceptance rate.
Emergency Ventilator
PID-controlled BVM compression system. Arduino Uno + stepper motor + MPX5010DP pressure sensor. Hardware watchdog timer — safety cutout is comparator circuit (not software). Build cost: ~₹3,200 (<$40) vs. ₹15L commercial. 8–22 BPM range, ±4% tidal volume accuracy across 500 cycles.
Tech Stack & Capabilities
Each skill is linked to a project where it shipped — no percentage bars.
HOVER = skill details · DRAG = physics interaction
Measurable Engineering Outcomes
Every number below is auditable — dataset, baseline, and constraint included.
LifeLineX — Real-Time Emergency Response
Emergency patients lose critical minutes because the nearest hospital may have no ICU capacity. LifeLineX connects patients, paramedics, and hospitals on a single WebSocket network — hospital availability is visible before ambulance arrival. Built with React, Node.js, Socket.io, MongoDB, and Twilio for SMS fallback.
Agri-Guard — CNN + IoT Crop Intelligence
15–25% of Indian crops are lost to disease annually because diagnosis is slow and expensive. Agri-Guard gives any smartphone user a 94.8% accurate crop disease diagnosis (MobileNetV2, PlantVillage dataset, 38 classes) and an ESP32+DHT11 IoT node for real-time field monitoring. Baseline comparison: Haar+SVM at 63%.
Vedic Asharam — Offline-First EdTech
Education platforms fail students with low bandwidth and visual impairment. Vedic Asharam uses Service Worker offline caching (cache-first for content, network-first for video), Web Speech API for audio, and was optimized for <2s First Contentful Paint on 3G. 42 structured lessons across 6 modules — no backend, $0 hosting cost.
The Timeline
From first-year student to multi-domain AI builder — tracked by what was shipped, not just studied.
Joined UEM Jaipur
Began B.Tech CSE at University of Engineering and Management, Jaipur — specializing in AI and cybersecurity.
First AI Projects
Built EmotiScan (CNN-based emotion recognition using OpenCV + Dlib) and an Emergency Ventilator prototype — combining hardware and software for social impact.
Scaling to Full-Stack AI
Shipped LifeLineX (live on Vercel) — a real-time emergency hospital network using Socket.io. Expanded into AgriTech with Agri-Guard's CNN + IoT dual system. Built Vedic Asharam for inclusive education.
Myntra Hacker Ramp Hackathon
Competed in India's top fashion-tech hackathon — built an AI-powered wardrobe intelligence platform for personalized fashion recommendations.
ML Deployment & Research
Deployed Customer Churn Predictor on HuggingFace Spaces. Built FoodLens AI nutrition API, Smart Retail AI dual-tool platform, and conducted defense-scale research on the Military Support System.
Open to Internships
Actively seeking internship and research opportunities in AI engineering, full-stack development, and applied machine learning. Ready to contribute immediately.
Peer-Reviewed IEEE Publications
3 published papers at the intersection of IoT, AI, and social impact — from smart farming to brain-computer interfaces.
Hackathons & Competitions
9 major competition results — including wins at Harvard and Stanford-level hackathons.
Harvard Hackathon
Stanford Hackathon
SRM Delhi Hackathon
HackUEM 2.0
Techfest UEM Jaipur
Skit Startup Expo — Peavah
AceHack 3.0
IIT Patna Global Coding Competition
Codeblitz
Let's Build Something
That Matters
Open to internship opportunities, research collaborations, and interesting projects across AI, full-stack development, and beyond. Response guaranteed within 48 hours.