M.Eng Autonomy and Robotics, University of Illinois Urbana-Champaign

Designed a low cost solution to digitally override a joystick control box of a grain probe using a custom PCB, ESP32 microcontroller and NVIDIA Jetson enabling automation of grain sample collection from a truck.

Developed and containerized a ROS 2 pipeline using Docker that detects incoming grain trucks using YOLO, builds a spatial map using Intel Depth Sense camera and controls the grain probe to take grain samples every 45 seconds.

Building a web based control interface using React, FastAPI and Uvicorn to configure the system, calibrate cameras, override operations, and start/stop the robotic pipeline.

Developed a real time product packaging assistant and validation system for faucets using a lightweight Mask RCNN instance segmentation model, achieving 30 ms inference latency on NVIDIA Jetson.

Generated a synthetic dataset to augment missing faucet color and finish variations, resulting in a model achieving above 90% accuracy and mAP on real world inspection.

Worked as a research assistant for Dr. Ben Blair, building AI training pipelines for swine wellness research.

Evaluated state of the art pose detection models trained on homogeneous and heterogeneous swine datasets to measure their impact on keypoint detection accuracy in real world farm environments.

Developed a web based application using React and Uvicorn to detect swine wellness through fecal matter classification.

Optimized multimodal deep learning ROS2 pipeline for STRETCH Robot AI, enabling compatibility with low end GPUs by reducing VRAM usage from 12GB to 7GB.

Improved the reliability and safety of elderly assistive object pick and place tasks, doubling the pickup success rate from 20% to 40%.

My course work includes working with mobile robotics and autonomous vehicles, currently focusing on developing my skill set on optimizing deep learning perception models and real time systems coordination.

Related Course Work: Autonomous Vehicle System Engineering, Deep Learning with Computer Vision, Computer Vision, Principles of Safe Autonomy

Designed an automated system to monitor and maintain Honeywell Lenel S2 security systems deployed in Microsoft Datacenters globally, ensuring badge readers, CCTV cameras, and security applications operated reliably, and triggering alerts when an edge device fails.

Evaluated performance of cognitive vision models by assessing accuracy in detecting and reporting faulty surveillance cameras across Microsoft data centers.

Developed Azure CI/CD pipelines to provision, configure, and manage Docker containers, Azure virtual machines, and BareMetal servers, reducing manual setup and maintenance effort by 60%.

Developed an IoT device to detect obstacles blocking access to fire extinguishers and alert security in real time for 5G Radio Chip Manufacturing Factory.

Built a Python Flask backend with SQL database to communicate with ESP32 microcontroller to continuously update sensor data and transmit real time status updates to the security system.

My course work included working with embedded systems, real time system processing and image processing

Related Course Work: Real Time Embedded Systems, Image Processing,