I build AI agents and the backends that keep them honest.

AI Engineer

Pattern Technologies — Pune, India

Working on automation workflows and backend systems.

Software Engineer

Kimaru.ai — Remote

Architected and built a production-grade supply chain optimization platform for a large carpet manufacturer, automating yarn requirement planning and generating optimized loom production schedules. Developed scalable full-stack services using FastAPI, Next.js, TypeScript, PostgreSQL, and Docker, and deployed them on AWS using containerized infrastructure.

Led a team of 5+ software engineering interns, owning technical architecture, feature planning, code reviews, and engineering best practices.

AI Engineer

Quark Inc — Japan

Delivered enterprise AI solutions for Japanese clients by building LLM-powered applications for knowledge management, document intelligence, and business process automation. Built an AI agent for automated 2D CAD-to-3D model conversion, combining LLM reasoning with geometric processing and human-in-the-loop validation to generate production-grade parametric 3D models.

Software Engineering Intern

Pattern Technologies — Pune, India

Built backend and fronted for AI-driven agentic workflows using LangGraph. Developed Image Auditing and Image Generation Pipeline to help generate high accuracy e-commerce product images. Worked with the team to create Apache Airflow DAG for various data pipelines.

Software Developer Intern

Kennemer Eco Solutions — Asia (Remote)

Built and maintained full-stack web applications using Django, React.js, SQL, and AWS. Developed scalable backend APIs, database models, and responsive dashboards for environmental sustainability solutions.

Set up AWS servers and CI/CD pipelines, scaling traffic to 3,000+ users/day.

AI Shopping Assistant

A voice-driven shopping assistant that runs a multi-step research pipeline over live retail data: it extracts intent from a natural-language query, plans the research, runs the search, then ranks results on price, rating, review sentiment and availability. Google Gemini 1.5 Flash handles the language understanding, SerpAPI supplies live Walmart listings, and the Web Speech API takes hands-free queries in Chrome and Edge. Also does review-sentiment analysis, brand and price-trend comparison, and a guided four-step checkout.

Snap2Scene

A comparative study of six ways to rebuild a 3D scene from 2D images, built as a computer vision course project: Gaussian Splatting — including semantic-driven and text-to-3D variants with an interactive WebGL viewer — incremental Structure-from-Motion, multi-view SfM, Neural Radiance Fields, Pix2Vox voxel prediction, and space carving. Implemented across Python notebooks and shipped as Snap2Scene, a live HuggingFace Space where the pipelines run in the browser.

Music Recommendation from Facial Emotion

A full-stack app that reads mood from a photo and recommends music to match it. A PyTorch and TensorFlow emotion-recognition model classifies the facial expression in an uploaded image, and the detected mood drives the track suggestions. React front end against an Express and Node API, with MongoDB behind account registration and login, and a layout that holds up on mobile.

Custom VGG16 and LeNet models trained on FER-2013, reaching 63.56% accuracy.

FoodBuddy

A food ordering platform where users browse dishes across multiple restaurants and order them end to end. Covers registration and login, price-range filtering so you can shop to a budget, a favourites list for saved dishes, and a cart through to checkout. Built as a React client against a Node API.

NutriNova

A diet tracking app that turns a fitness goal into a daily plan you can actually follow. You enter your metrics — weight, height, age, activity level — pick a target, then log meals as you eat them, and NutriNova scores the nutrition against that goal and adapts its recommendations as your intake shifts. Progress is charted over time, and a built-in LangChain chatbot answers questions about your plan, your logged meals and what to eat next. React front end, Node API and MongoDB, with accounts so history persists across devices.