SYS:// VISION_PIPELINE.v2
MODEL:// YOLOv8 · SAM · ResNet
STATUS:// ● ONLINE
LOC:// PUNE, IN

Teaching machines to see,
so humans don't have to look twice.

$

I'm Rishiraj Behera, an M.Tech grad in Computer Science building perception systems that cut inspection time and catch defects humans miss. Worked as an Analyst Trainee at Deloitte, now working at the intersection of deep learning, manufacturing, and real-time video at Tata Motors.

03 INDUSTRY PROJECTSPROJECTS
05 PROJECTS SHIPPEDSHIPPED
2025 M.TECH · NIT ROURKELAM.TECH · NITR
01 //

About

Rishiraj Behera
EDGES · 0.99ID: RISHIRAJ_B

I build systems that let cameras replace slow, subjective quality checks on factory floors — detection, segmentation, tracking, and the glue code to push decisions into production pipelines in real time.

I'm currently a Graduate Apprentice Trainee at Tata Motors Digital.AI Labs, working on computer vision for quality inspection. The job is straightforward to describe and hard to do well: teach a model to spot defects faster and more consistently than a trained human eye, then earn engineers' trust enough to act on it.

Before this, I worked full-time as an Analyst Trainee at Deloitte USI (Node.js, microservices, REST APIs) — which turns out to be an underrated advantage: real ML systems are 10% model and 90% plumbing.

// Recently completed M.Tech in Software Engineering from NIT Rourkela (CGPA 8.17), with a thesis in deep learning for biometric identification.

Role
GAT @ Tata Motors AI Labs
Stack focus
PyTorch · OpenCV · YOLO · SAM
Education
M.Tech, NIT Rourkela · 2025
Open to
CV / ML roles, collaborations
02 //

My Works

Animal Biometrics
CATTLE · 0.97
MUZZLE
01 · COMPUTER VISION · PYTORCH

Animal Biometrics

Deep-learning model for individual cattle identification from muzzle images — a contact-free alternative to RFID tags. Trained a custom CNN with data augmentation to handle real-world noise: lighting, occlusion, animal motion.

PyTorchOpenCVCNNAugmentation
Concurrent Web Server
SYSTEM · C
02 · SYSTEMS · C

High-Performance Concurrent Web Server

Multi-threaded HTTP server in C — thread pools, epoll-based I/O, graceful shutdown. Underpins why I'm comfortable debugging latency in ML serving stacks today: the bottleneck is almost never the model.

CPthreadsepollHTTP
Hostel Management System
MERN APP
03 · FULL-STACK · MERN

Hostel Management System

End-to-end MERN app for managing hostel occupancy, complaints, and fee cycles. Role-based auth, admin dashboard, responsive UI. Built to actually ship, not just demo — used for real record-keeping.

ReactNode.jsExpressMongoDB
Lazy Reader
Q&A PLATFORM
04 · FULL-STACK · MERN

Lazy Reader

A Quora-style Q&A platform with feeds, voting, comments, and topic-based discovery. JWT auth, rich-text posts, and the kind of ugly-but-honest state management you only learn by building your own social feed from scratch.

ReactNode.jsJWTMongoDB
Maverick-007 Game
SPACESHIP · 0.94
05 · GAME · VANILLA JS

Maverick-007

2D space shooter in vanilla JS — collision detection, particle effects, scoring, and a HUD that I definitely over-engineered. Deployed on Netlify. Prototype muscle that still shows up every time I sketch a proof-of-concept.

JavaScriptCanvasNetlify
03 //

Stack

Computer Vision & ML

/01
PyTorchOpenCVYOLOv8SAM ResNetCNNDetectionSegmentation TrackingONNX

Languages

/02
PythonCJavaScriptTypeScript SQLBash

Full-Stack & Backend

/03
Node.jsExpressReactREST APIs MongoDBMySQLMicroservices

Tooling

/04
GitDockerLinuxVS Code JupyterPostman
04 //

Contact

Let's build something that sees.

Open to CV / ML engineering roles, research collaborations, and interesting problems where a camera can make a decision. Drop a line — I reply fast.