M.Tech Software Engineering student at VIT Vellore specialising in backend systems and applied ML. Author of 2 published patents in cryptographic authentication and federated learning. Holds AWS ML Engineer and Solutions Architect certifications. Active open source enthusiast — contributor to production codebases used by tens of thousands of developers.
3Projects
4+OSS Repos
2Patents
5+Certifications
8.07CGPA
Python & Java
FastAPI · REST APIs
Retrieval-Augmented Generation
YOLOv8 · Computer Vision
Docker & Firebase
AWS Cloud Practitioner
Open Source Contributor
VIT Vellore
Python & Java
FastAPI · REST APIs
Retrieval-Augmented Generation
YOLOv8 · Computer Vision
Docker & Firebase
AWS Cloud Practitioner
Open Source Contributor
VIT Vellore
Skills
Technical Stack
Languages
Java
Python
JavaScript
SQL
Backend
FastAPI
Node.js
REST APIs
Authentication
SSE Streaming
Databases
SQLite
Firebase
Supabase
AI / ML
RAG Pipelines
YOLOv8
TabNet
Scikit-learn
NLP
Cloud & DevOps
AWS (EC2, S3, IAM)
Docker
Git
GitHub
Linux
Data & Retrieval
ChromaDB
FAISS
Pandas
NumPy
CS Fundamentals
Data Structures & Algorithms
OOP
Computer Networks
Operating Systems
HTTP / TCP-IP
Engineering Work
Selected Projects
01
StructRAG: Hybrid Tree-Boosted RAG for Academic Document Analysis
Builds a hierarchical section tree during ingestion using a 13-type academic ontology. ATB-RRF algorithm fuses vector similarity with structural signals at query time — achieving 23–41% better retrieval over vector-only RAG at just 3ms extra latency. Powered by ChromaDB, Gemini Flash, and streaming SSE responses with citation highlighting.
AI-powered early detection of sugarcane pests using YOLOv8 for image detection and TabNet for tabular prediction. Containerised with Docker, served via FastAPI.
Android ride-sharing app with real-time ride posting, Firebase Authentication, Realtime Database, and Google Maps SDK for route visualisation and navigation.
Fixed indented list continuation bug in TypeScript (composebox_typeahead.ts) — correctly preserves indentation when continuing indented markdown lists in the message composer.
LiteLLM retry with exponential backoff, zero-shot MNLI safety guardrails, SQLAlchemy type mapping, wearable sensor data validation, and test teardown connection cleanup.
Fixed the scheduler UI to display same-classroom sections in the section info panel. Added unit tests for the LotteryFrontendModule ensuring robust scheduling logic.
