Umeshkumar Ghaskata

Software Engineer at Roswell Park Comprehensive Cancer Center (2025-06 – Present)

Owned the full software development lifecycle(SDLC) for a cross-platform React Native application using JavaScript, Firebase, and SQL-based data concepts for behavioral tracking and personalized AI coaching by integrated REST APIs.
Managed source code and collaborative development using Git and GitHub, implemented automated CI/CD pipelines with GitHub Actions, and improved application performance by reducing redundant API calls using Redux by 30%.
Built personalized AI coaching by Fine-tuned LLM on AWS Cloud Linux instance using Python, PyTorch. Then deployed it to a university server to eliminate reliance on OpenAI API calls and reduced on going maintenance cost.
Implemented Redis caching and rate limiting in a Node.js Express backend, reducing redundant AI API requests and improving backend performance and response latency.
Followed Agile practices through weekly meetings with user, advisor: taking direction and constructive criticism.

Software Engineer at Scitara Corporation (2022-11 – 2023-02)

Build and deployed low latency RESTful API for SaaS Product using Docker which reduced the manual testing time by 30%, removed the reliance of 3rd Party API for testing.
Facilitate backend with FastAPI Microservices with JWT Authentication and MockAPI service with PostgreSQL, equipped with Kubernetes(AWS EKS) for production.

Volunteer Web Developer at IIT Bombay - Rocket Team (2022-03 – 2022-07)

Developed and maintained the IIT Bombay Rocket Team's multi-page static website using HTML, Tailwind CSS, and JavaScript, implementing responsive UI components and deploying the application through GitHub Pages.

Research Assistant at DRONES Lab (NASA AIST-21 Project) (2024-08 – Present)

Designed, trained, evaluated, and deployed Deep Learning models (PCA+MLP, FPGA-based architectures) for hyperspectral image classification improving land-cover recognition accuracy from 75.4% to 94.5%.
Reduced edge inference latency from 295ms to 52ms using optimized Model to enable real-time classification.
Engineered benchmarking framework across RTX 4060, Jetson Orin, and Xavier guided Hardware selection.
Published Thesis on energy-efficient coverage path planning; developed a Python-based optimization framework that processed millions of simulation data reducing path overlap by 40%, improving energy efficiency by 10%.

About