Hi, I'm AI Software Engineer
Umesh Ghaskata
AI Software Engineer specializing in Generative AI, LLMs, RAG, and Agentic AI. M.S. Computer Science graduate from the University at Buffalo and IIT Bombay alum. I build production-ready intelligent systems that bridge research and real-world deployment.
About Me
I'm Umesh Ghaskata, an AI Software Engineer and Computer Science graduate from the University at Buffalo, previously at IIT Bombay. I'm passionate about building production-ready intelligent systems that bridge research and real-world deployment.
My work spans Generative AI, LLMs, multimodal RAG, agentic AI, and cloud-native software engineering. At the DRONES Lab, I contributed to NASA AIST-21 research on hyperspectral image classification and UAV path planning, reducing edge inference latency from 2.7 seconds to 32 milliseconds through FPGA-based optimization.
Beyond research, I've built end-to-end systems including multimodal RAG platforms, LLM fine-tuning pipelines (LoRA/QLoRA), GenAI-powered web applications, and cross-platform mobile apps. I'm experienced with Python, PyTorch, FastAPI, AWS, React Native, vector search, and MLOps workflows.
Portfolio: umeshghaskata.com
Education
University at Buffalo, SUNY
Master of Science (Computer Science)
2024 – 2026 | Buffalo, NY | GPA: 3.86/4.0
Distributed Robotics and Networked Embedded Systems (DRONES) Lab
IIT Bombay
Bachelor of Technology (Aerospace Engineering)
2020 – 2024 | Mumbai, India
Experience
Full Stack Engineer
Roswell Park Comprehensive Cancer Center · Part-time · Jul 2025 – May 2026 · Buffalo, NY
- Developed cross-platform mobile applications using React Native 0.78 with multi-screen navigation, daily check-ins, push notifications, and local storage for iOS and Android.
- Fine-tuned Gemma 3 1B-IT on AWS EC2 GPU using QLoRA, Hugging Face PEFT, LangChain, and LlamaIndex; deployed a personalized chatbot on a UB server.
- Engineered CI/CD pipelines reducing deployment time by 40% and optimized Redux state management to reduce redundant API calls.
Student Assistant | DRONES Lab
The Research Foundation for SUNY · Part-time · Aug 2024 – Apr 2026 · Buffalo, NY
- NASA AIST-21: Designed and deployed DL models (PCA+MLP, FPGA architectures) for 274-band hyperspectral image classification, achieving 82.1% accuracy while reducing inference latency from 2.7s to 32ms for real-time edge deployment.
- Integrated a UAV platform with Jetson Orin, Xavier, ULTRIS-X20 sensors, and CUDA/TensorRT acceleration for onboard edge-AI inference.
- Built a cross-platform benchmarking framework across RTX 4060, Jetson Orin, and Xavier on self-annotated ULTRIS-X20 samples.
- Formulated an optimization algorithm for efficient disk packing in obstacle-rich areas using shape derivatives and LKH-D motion planning, reducing overlap by 45% and energy consumption by 10–12%.
Software Engineer Intern
Scitara Corporation · Full-time · Dec 2022 – Feb 2023 · Mumbai, India
- Reduced manual testing effort by 30% by implementing virtual server responses to mock RESTful API responses using Mockoon, enabling automated validation of GET/POST requests.
Web Developer
IITB Rocket Team · Full-time · Mar 2022 – Jul 2022 · IIT Bombay
- Built and maintained the team website using HTML, CSS, and responsive layout techniques.
Projects
Multimodal RAG Platform
Personal Project |
Built a production-ready Multimodal RAG platform that answers questions from PDFs, DOCX files, and images using semantic search and LLMs. Developed an end-to-end ingestion pipeline with OCR, image extraction, AI captioning, intelligent chunking, embedding generation, and vector indexing. Implemented context-aware retrieval using OpenAI Embeddings and Supabase pgvector with cosine similarity search, enabling accurate responses with document citations.
View Code Watch DemoSupervised Fine-Tuning Gemma 3-1B LLM with LoRA vs QLoRA
Personal Project |
Experimented with parameter-efficient fine-tuning (PEFT) by training Gemma 3 1B IT on a custom Smoking Cessation / Motivational Coaching dataset (~1,500 conversations) on AWS EC2 (Tesla T4). Compared LoRA and QLoRA (4-bit NF4 quantization) trade-offs using BERTScore and ROUGE-L metrics. Both methods significantly outperformed the base model.
View Code View Model
GenAI-Powered Career Readiness Platform
Personal Project |
Built a career preparation platform that uses Generative AI to analyze resumes against job descriptions, identify skill gaps, and generate personalized interview preparation plans. Leveraged Google Gemini to create ATS-friendly resume content, technical/behavioral interview questions, candidate-job match scores, and structured learning roadmaps.
Tech Stack: React.js, Node.js, Express.js, MongoDB, Google Gemini, JWT, Zod, Axios, Puppeteer, PDF-Parse.
View CodeImage-Caption-Generator
University at Buffalo |
End-to-end Deep Learning-based Image Caption Generator that automatically generates natural language descriptions for images. Combines Computer Vision and NLP using a pre-trained CNN for feature extraction and an LSTM-based sequence model for caption generation, learning to produce meaningful captions word-by-word from input images.
View Code
Autonomous Drone for Search Operations
IIT Bombay |
Developed an autonomous drone using Pixhawk + RPi, HOG-based detection, and A* algorithm with ultrasonic obstacle avoidance.
View CodeML Projects: Logistic Regression, CNNs & RL
SUNY Buffalo |
Implemented logistic regression from scratch using gradient descent and L2 regularization. Built and tuned CNNs and MLPs in PyTorch with dropout, batch normalization, LR scheduling. Designed a custom RL environment and trained agents using SARSA and Double Q-learning.
View Code
Bollywood Celebrity Recognition
IIT Bombay |
Built a facial recognition app using VGGFace (ResNet50) and MTCNN for face detection and embedding. Stored 2048-feature vectors using Pickle. Deployed a web app using Streamlit for real-time celebrity classification.
View App
Coordinated UAVs for Efficient Spraying
IIT Bombay |
Modeled a 2D grid-based spraying path using OpenCV and HSV heatmaps. Applied Traveling Salesman Problem to optimize waypoint sequence and simulated A* and Dijkstra algorithms for flight efficiency.
View Report
Operating Systems Programming
IIT Bombay |
Built a Linux shell using fork(), exec(), wait() and implemented dynamic memory management in xv6. Used pthreads, mutexes, and semaphores for multi-threaded synchronization in user-space programs.
View Code