CV

Computer Vision Engineer / Research Assistant

Institute for Simulation and Training (IST)

• Conducting research and prototyping for embedded computer vision and visual navigation systems, with a focus on robust perception and multi-sensor localization.
• Tech stack: ROS 2, PyTorch, XFeat, visual-inertial SLAM, ONNX Runtime, TensorRT, and NVIDIA Jetson.

Software Engineering Intern

Xometry

• Developing an AWS Bedrock Claude SEO agent supporting a pilot project reporting +50% quotes, +30% sessions, and +86% AI-referral CVR; planned 2 QA/monitoring agents to scale JSON-LD across 1,200+ Xometry pages and content.
• Built an object-oriented Python CLI that fetches content from Prismic and uses MongoDB/PyMongo for persistent storage, while skipping unchanged pages to prevent unnecessary Bedrock calls.

Software Engineering Fellow (Certified)

Bloomberg Tech Lab

• Selected for Bloomberg Tech Lab and collaborated with a teammate to build a real-time messaging system using RabbitMQ.
• Developed the consumer component in Python, configured message queues, and ensured reliable message processing and acknowledgment.
• Used Git and Docker for collaboration and deployment, tested the system end-to-end, and delivered a fully functional solution on time.

Software Developer Intern

Prodisphere

• Worked on an athlete evaluation algorithm using performance stats, social media data, and experience to support athlete-brand matching.
• Implemented a weighted scoring system to generate overall athlete ratings for NIL (Name, Image, Likeness) opportunities.

Undergraduate Teaching Assistant

University of Central Florida

• Assisted Lecturer Dr. Tanvir Ahmed with a Data Structures and Algorithms course, supporting 70+ undergraduate students for 10 hours per week.
• Contributed to a 15% increase in the average passing score for students in one of UCF’s most demanding computer science courses.

Software Engineer Intern

UrbanITY Lab

• Developed real-time LiDAR data visualization pipelines using object-oriented C++ to improve computer vision and perception systems for autonomous vehicles.
• Designed and implemented statistical and deep learning algorithms for background filtering, object-of-interest clustering, and bounding-box regression to enhance object detection accuracy.
• Presented technical findings to research stakeholders, translating computer vision results into actionable engineering updates.

Master's

University of Central Florida

Computer Vision

AI Fast Food Ordering Agent

A system that takes voice or text commands, such as “Order me a cheeseburger,” and automatically finds the nearest restaurant to start the order.

• Created a reliable scripted pipeline that extracts order details from an LLM and uses Python scripts to navigate sites like McDonald’s, ensuring 100% order accuracy.
• Architected the deployment using Docker and NGINX to host the application on a high-performance A100 GPU server, secured with SSL for encrypted browser access.

Model Context Protocol (MCP) with Google Calendar Integration

A Node.js server using the Model Context Protocol (MCP) to provide AI applications with direct access to Google Calendar data.

• Enabled AI agents to fetch calendar events by date and return formatted meeting information through a standardized interface.

Vision Transformer (ViT)-Based ASL Recognition

An American Sign Language classifier built by fine-tuning Hugging Face’s pretrained Vision Transformer (ViT), using data augmentation and class weighting.

• Achieved an overall accuracy of 94% on a diverse dataset of over 26,000 images.

Pokemon GO-Inspired Web App

A Pokemon GO-inspired web application using the MERN stack, integrating the Pokemon API to enable users to guess Pokemon names from images.

• Engaged more than 200 players, demonstrating the impact of the web application within the community.

Quantize the Language Tower, Not the Vision Tower: Recovering Numeric Accuracy in Page-Image QA

Manuscript submitted to IEEE FMLDS 2026.

User Profile with Large Language Models: Construction, Updating, and Benchmarking

arXiv preprint arXiv:2502.10660.

PEFT A2Z: Parameter-Efficient Fine-Tuning Survey for Large Language and Vision Models

arXiv preprint arXiv:2504.14117.