Building AI-powered systems and production-grade software — from computer vision pipelines and explainable ML to role-aware full-stack platforms.
I'm a final-year CS student who builds things at the intersection of ML and software engineering. My work spans transfer learning models, full-stack academic platforms with explainable AI, sustainability-focused AI tools built for real internships, and parallel computing experiments in C++.
What I care about: systems that are honest about uncertainty, code that's structured enough for someone else to run without a phone call, and ML paired with real interfaces — not just notebooks.
Currently seeking internships in AI/ML engineering, full-stack development, or data science — December 2026 onwards.
AI / ML → Computer vision · Transfer learning · Explainable AI (SHAP) · Rule-based classifiers
Full-Stack → React 19 + TypeScript + Supabase · Role-based access · Real-time dashboards
Sustainability → AI for waste segregation · Air quality forecasting (SDG-aligned)
Systems → Parallel computing · Dense matrix multiplication in C++
Learning Now → LLMOps · RAG pipelines · FastAPI · Agentic AI
|
Academic Management Platform Role-aware portal for engineering colleges with 4 dashboards — Student, Faculty, HOD, Admin — each scoped by RS256 JWT and Supabase Row Level Security. The AI layer runs XGBoost + Random Forest on attendance and marks data and surfaces at-risk student alerts in the faculty view. Each alert includes SHAP explainability so faculty understand why a student was flagged.
|
♻️ AI-EcoSortAI-Powered Waste Segregation Assistant Built for the 1M1B IBM SkillsBuild AI for Sustainability internship (SDG 12). Classifies waste into 5 categories with disposal guidance. Runs fully offline via a local rule-based classifier — no database, no external API required. Optionally connects to IBM Granite via watsonx.ai. Ships with unit tests, Architecture docs, Responsible AI plan, Privacy policy, and a confidence meter that communicates uncertainty honestly.
|
|
Computer Vision — Transfer Learning EfficientNetB0 transfer learning model for waste image classification across 6 categories.
Ships with a Gradio demo app and full retraining instructions.
|
Financial Calculator Portal · Live → SIP, EMI, and goal-based financial calculators with a clean, responsive UI. Deployed on Vercel.
🛒 AeroBuyE-Commerce Backend Flask e-commerce with product catalog, cart, checkout, and MongoDB data layer. Session management via Flask-Login.
|
|
Parallel Computing · Performance Engineering Parallelisation strategies for dense matrix multiplication in C++. Includes full project report covering algorithm design, benchmarking methodology, and parallelisation overhead analysis.
|
🌍 Atmos (private)Air Quality Forecasting Multi-source ETL pipeline and time-series forecasting for air quality data. 88% accuracy, built to handle 500+ daily data requests.
|
| Domain | Technologies |
|---|---|
| Languages | Python · TypeScript · JavaScript · C++ · SQL |
| AI / ML | TensorFlow · scikit-learn · XGBoost · SHAP · EfficientNet · Gradio · IBM Granite |
| Frontend | React 19 · Vite · Tailwind CSS · Vitest |
| Backend | Flask · Supabase · PostgreSQL · MongoDB |
| Tooling | Docker · Git · Oxlint |
| Learning | FastAPI · RAG · LLMOps · Agentic AI |
- 🔭 Building: RAG pipelines and FastAPI services
- 🌱 Learning: LLMOps, agentic AI workflows, MLOps tooling
- 🎯 Open to: Internships (Dec 2026) in AI/ML Engineering · Full-Stack · Data Science
- 🏅 Won: NASA Space Apps Challenge
"Build things that are honest about what they can and cannot do."