Parth Gupta

1. Graduate Research Associate (Software Engineering & AI Systems)

   The Ohio State University — NOWLAB

   Aug 2024 — May 2026 · Columbus, OH

   • Graduate Research Associate in NOWLAB, working with Dr. Hari Subramoni and Dr. D.K. Panda on AI-powered digital agriculture within the NSF-funded ICICLE AI Institute — turning drone and field imagery into actionable insights via computer vision, geospatial analytics, and HPC.
   • Built and maintained Harvest, an NSF-funded drone imagery analysis platform, recognized as “Most Innovative Software” at Ohio’s Farm Science Review 2025 (also referred to as the ICICLE digital agriculture pipeline).
   • Developed the Harvest web app (Python/Flask + React + PostgreSQL) for job submission, GPU job monitoring, result visualization, and secure downloads — reducing data-access time for growers and researchers while cutting shapefile output size by ~60% and map rendering time by ~40%.
   • Designed and shipped a vision transformer–based computer vision stack in PyTorch that converts raw drone field imagery into structured JSON detections, improving crop/weed detection accuracy and simplifying downstream analytics tools.
   • Developed production RAG systems with Chroma/Pinecone and fine-tuned open-source LLMs (Mistral, Zephyr) via LoRA/QLoRA, improving retrieval precision by ~28% and factual consistency by ~45% on internal 1000+ documents.
   • Engineered DL-based crop inference and classification with spray-zone shapefile export for targeted operations; containerized the stack and deployed via Docker + Tapis on Ohio Supercomputer Center resources (and integrated paths for large-scale runs including San Diego Supercomputer Center workflows where applicable).
   • Profiled and optimized GPU inference using CUDA profiling, batch-size tuning, pinned memory, async data transfers, transfer–compute overlap, and GPU-side preprocessing — cutting per-image runtime from ~0.86 s to ~0.19 s (4.5× faster) to enable near real-time field-scale analysis.

2. Volunteer Software Engineer

   APenergy

   Jan 2025 — Mar 2026 · Columbus, OH

   • Built an internal live project dashboard (React, PostgreSQL) to track project status, timelines, and blockers; improved visibility into bottlenecks for faster decisions.
   • Designed and shipped Excel-based lead tracking with a lightweight website (Django, PostgreSQL) that automates follow-ups and organizes leads.
   • Developed a RAG chatbot (LangChain, OpenAI API, Pinecone) that answered website visitor questions using company docs and policies, handling ~30% of inbound visitors.

3. Software Engineer

   Webonex Technologies LLP

   Jun 2023 — Oct 2023 · Indore, India

   • Implemented backend REST APIs using Java and Spring Boot, including OAuth 2.0 authentication, enabling secure user access to the platform.
   • Optimized SQL queries and database indexing, reducing average response time by ~30–40 ms across high-frequency endpoints and improving throughput.

4. Computer Vision Engineer (Research Intern)

   Indian Institute of Technology Indore

   Feb 2023 — May 2023 · Indore, Madhya Pradesh, India

   • Built computer vision pipelines for high-resolution satellite imagery preprocessing and feature extraction using Python, OpenCV, and GDAL — improving classification and segmentation accuracy by ~18% and reducing processing time by ~0.23 s/tile via efficient data structures and algorithms.

• Monetizer.ai↗

  Public demo where users chat with an AI in a creator’s voice, grounded strictly in that creator’s transcript/caption data (not generic web knowledge), with optional affiliate answers and voice UX.

  Flask, HTML/CSS/JS, Supabase, sentence-transformers (all-MiniLM-L6-v2), RAG with RPC vector retrieval, cosine-similarity fallbacks, relevance thresholds

• Go Rent↗

  Secure vehicle-rental backend with user records and OTP verification via Twilio; recognized as the Most Innovative Project of the University.

  Java, Spring Boot, Twilio, secure flows and validation

• Potato Disease Classification↗

  Web app serving a CNN to classify potato leaf diseases into three categories with ~90% accuracy; reduced false negatives by ~23% through preprocessing improvements for field reliability.

  CNN deployment, augmentation, normalization, Flask web serving

• Health Record↗

  Person identification through a webcam-enabled site: CNN-based recognition, then medical history on screen with Google Text-to-Speech readout for accessibility — combines computer vision, deep learning, and NLP.

  Python, CNNs, Streamlit, Google Text-to-Speech, webcam pipeline

• Banking website (JSP)↗

  Early academic project: full-stack banking site for accounts, inter-bank transfers, and transaction history — JDBC/MySQL backend and a Bootstrap + JSP/HTML/CSS UI.

  Java, JSP, Servlets, JDBC, MySQL, HTML, CSS, Bootstrap

• Gupta, P., Gopinath, H. B., Patil, V., Dean, C., Potlapally, A., Subramoni, H., Shearer, S., Venkatesh, R., Fulton, J., and Stewart, C. Auto-Harvest: An End-to-End Automated Cyberinfrastructure for Precision Agriculture Workflows. ASABE Annual International Meeting, 2026, Paper No. 2600653.

  Conference paper on the Harvest precision-agriculture cyberinfrastructure platform.

• Gupta, P. UAV-Based Weed Detection and Spot-Spraying Prescription Generation: A Scalable Computer Vision and Geospatial Framework. Master’s thesis, The Ohio State University. OhioLINK Electronic Theses and Dissertations Center.

  Approved and accepted by The Ohio State University; in process of publication.

  View on OhioLINK →

I’m happy to connect about software engineering, ML systems, or full-stack work. Reach out by email or phone, or find me on GitHub and LinkedIn.