Software Engineer with an M.S. in Computer Science from The Ohio State University (May 2026), based in Columbus, OH. I build systems where machine learning, solid backend engineering, and deployment intersect, from profiling and performance tuning to shipping usable interfaces.

At OSU NOWLAB / ICICLE, I worked on AI for digital agriculture: computer vision on drone and field imagery, GPU optimized inference, geospatial outputs (spray zone workflows), and React + Flask tooling, with pipelines run on HPC using Docker, Tapis, and OSC. Beyond computer vision, I have built LLM powered applications with RAG pipelines, combining retrieval, embeddings, and vector databases to ground language model responses in domain specific knowledge. I have also designed and implemented multi agent orchestration systems where specialized agents collaborate to produce stronger outputs than a single prompt chain allows. I like owning problems end to end, from system architecture and model performance to a usable product interface.

• ML in production — vision models, training and inference workflows, faster pipelines
• LLMs and RAG — retrieval, embeddings, pragmatic product constraints
• Backend and full-stack — Java, Python, APIs, databases, shipping reliable services
• Systems and scale — Linux, containers, distributed and high-throughput workloads

End-to-end AI pipeline for agricultural drone autonomy. Integrates waypoint-based mission planning, real-time object detection and classification (YOLO, OWLv2), geospatial shapefile generation, and a React + Flask interface. Runs on OSC HPC via Docker and Tapis, built in collaboration with ICICLE, the NSF-funded AI institute for intelligent cyberinfrastructure.

A RAG-powered chatbot that lets fans chat with a creator's knowledge base built from their video transcripts and captions. The pipeline ingests creator content, extracts and embeds transcripts, and stores them for retrieval — so the assistant answers from the creator's actual words rather than generic web knowledge. Replies can also surface affiliate links for creator monetization.

Stack: Groq (LLM), Supabase (vector store), sentence-transformers (embeddings), Python backend

Highlights: Full ingestion, embedding, retrieval, and response pipeline with optional voice support; affiliate link injection configurable via environment variables

A local web application for personalized cold email outreach. Upload a contact list, compose once manually or with AI assistance, review every draft, and send through your own Gmail or Outlook account. Implements multi-agent orchestration where Agent 1 identifies resume-to-company alignment and Agent 2 writes the email from that research, producing stronger, more targeted drafts. Everything runs locally — spreadsheets, drafts, and logs stay on your machine.

Stack: Python, Groq and Gemini (AI drafts), Excel and CSV parsing, Gmail and Outlook SMTP

Highlights: Multi-agent orchestration for AI-generated emails; company research via public About page scraping; per-contact review and edit before send; rate limiting built in at approximately 50 emails per day

A deep learning web application that classifies potato leaf diseases from images. Built to help farmers and agricultural researchers get fast, accurate diagnoses without requiring technical expertise.

Stack: Python, convolutional neural network, web frontend

Highlights: End-to-end image classification pipeline; clean inference interface designed for non-technical users in agricultural settings

A Java-based vehicle rental platform built on MVC architecture using JSP, Servlets, and MySQL. Manages the full rental lifecycle from user registration to automated billing and invoice generation.

Stack: Java, JSP, Servlets, MySQL

Highlights: OTP-based verification for secure transactions; automated billing calculated from meter readings and rental duration; admin verification workflows for registration; role-based flows for customers and administrators

• Email: haeparth@gmail.com
• LinkedIn: parth-gupta-7a1685196
• Portfolio: parth-portfolio-ivory-gamma.vercel.app

Always happy to connect about software, ML systems, or collaboration.