Computer Vision · Multi-Agent AI · Papaya Orange at Heart 🧡
class Vansh:
team = "Papaya Orange 🧡 (McLaren supporter, rain or shine)"
lane = "Computer Vision 👁️"
day_job = "B.Tech CSE (AI & ML) @ SRM IST · Class of 2027"
based_in = "Chennai, India"
last_stint = "AI/ML Intern @ Tata Steel (Jan–May 2026)"
on_track = ["PlacementOS", "multi-agent systems", "RAG pipelines"]
sundays = "Race day. Strategy calls. Strong opinions about tyre choice."
off_circuit = "🎵 singing"
open_to = ["internships", "full-time roles", "research collabs"]| PRECISION rebar inspection |
LAP TIME manual inspection |
TOP SPEED in-browser face tracking |
CARS ON TRACK autonomous agents |
SENSORS river stations ingested |
I like problems where a camera and a model replace a human squinting at something: steel bars on a factory floor, a driver's eyelids at 100 km/h, litter on a city street, an empty parking bay. The recipe is usually the same.
| OpenCV rib measurement & ring-test analysis on live production lines | 478 landmarks, EAR / MAR / head pose, all client-side | YOLOv8 + DCGAN augmentation for urban cleanliness scoring | Overhead-camera parking occupancy in real time |
AI Career Command Center A local-first platform running 9 autonomous agents: job discovery, eligibility checks, JD-aware resume tailoring and application tracking. A RAG pipeline reads the job description; a TensorFlow model predicts salary bands.
|
Production computer vision (internship) Automated QC for TMT rebar: rib parameters measured at 95% precision, plus a spatial-analysis module for chemical ring tests that cut inspection time 40% across 3 production lines.
|
🌊 AERISFlood Risk Intelligence for Bihar 7-day flood forecasts from dual XGBoost models fed by 18 river stations, with a live React-Leaflet GIS dashboard and automated breach prediction for evacuation planning.
|
Driver Safety Vision System 478 3D facial landmarks at 60 FPS via MediaPipe. Microsleeps, yawns and distraction are classified from EAR, MAR and head pose, entirely in the browser with voice alerts. Zero server cost.
|
| Pos | Project | What it does | Stack |
|---|---|---|---|
| P5 | 🌍 QuakeIntel | Global seismic hazard intelligence with a volumetric 3D tectonic globe | XGBoost Three.js DBSCAN |
| P6 | 🗑️ LitterVision | Urban Cleanliness Index from municipal imagery, edge-ready | YOLOv8 MobileNetV2 DCGAN |
| P7 | 📄 Nexus-RAG | Upload PDFs, then interrogate them with semantic search + LLM reasoning | RAG LLM |
| P8 | 📡 Voxera | Low-latency P2P voice, video and screen sharing | WebRTC Node.js Socket.io |
| P9 | Real-time parking occupancy from overhead camera feeds | OpenCV Python |
|
| P10 | 🍽️ MealSync | Fair dinner decisions for flatmates via collaborative filtering | FastAPI XGBoost |
👁️ = computer vision project
2026 Jan–May AI/ML Intern · Tata Steel, Chennai
│ └─ Computer vision QC for TMT rebar · 95% precision · −40% inspection time
│
2024 Dec–Jan Internship Trainee · Digital Info Solutions, Ghaziabad
│ └─ Flask + MySQL full-stack apps · 3+ client projects delivered
│
2023 ─ 2027 B.Tech CSE (AI & ML) · SRM IST, Kattankulathur
└─ Currently on the final stint 🏁
 Fast, pushing hard right now Computer vision · OpenCV · MediaPipe · YOLOv8 Multi-agent orchestration · RAG pipelines |
 Reliable race pace FastAPI · React · XGBoost · TensorFlow · PyTorch ChromaDB · LangChain / LangGraph |
 Built to last the distance Python · SQL · JavaScript / TypeScript C/C++ · Java · Git · Docker |
"Box, box." Translation: I'm open to internships, full-time roles and research collaborations in computer vision, multi-agent systems and applied ML. Bring a hard problem and a camera.
🎵 Off the circuit I sing. Fast cars and good systems both come down to structure that lets something breathe.
🏁 Chequered flag. Thanks for stopping by the pit lane.