Technical Projects

Selected engineering, machine learning, robotics, and research projects
with emphasis on systems I designed, built, tested, and improved

View Project

Tags: Verilog · Python · Zynq-7000 · ARM Cortex-A9 · AXI-Lite · CNN · Quantization

Real-time privacy filtering before video storage or transmission

FPGA-Based Embedded Vision System

Description:
Built an on-device computer vision system that detects privacy-sensitive visual content and triggers selective Gaussian blur masking in real time.

Highlights:
· Designed a hybrid ARM/FPGA architecture for live video processing.
· Used sliding-window CNN inference to produce bounding boxes.
· Optimized a lightweight 32×32 CNN to 26 KB and reached an estimated 6 FPS.

View Project

Tags: C++ · EMG · Embedded Systems · Signal Processing · 3D Printing · Hardware Integration

An EMG-controlled prosthetic arm designed for more
accessible prosthetic technology

Emerging Prosthetics

Description:
Designed and built a muscle-signal-controlled prosthetic arm using embedded systems, real-time signal processing, 3D-printed parts, and custom hardware integration.

Highlights:
· Built and tested multiple working prototypes.

· Improved mechanical robustness by 70% and responsiveness by 40%.

· Presented the project at four conferences/fairs and earned national awards.

Tags: AutoCAD · Mechanical Design · Vehicle Systems · Prototyping · Team Engineering

Chassis and steering design for an autonomous electric vehicle team.

Princeton Autonomous Vehicle Electric

Description:
Designed chassis and steering components for Princeton’s autonomous vehicle team, supporting mechanical integration and collaboration with software developers.

Highlights:
· Created 3D CAD designs for chassis and steering components.
· Worked within motion, integration, and prototyping constraints.

· Collaborated across mechanical and software subteams.

Tags: Python · MATLAB · Statistical Analysis · Surveys · Healthcare Research

Quantitative research on healthcare algorithm
accuracy and online medical information.

Digital Diagnoses

Description:
Conducted statistical analysis on healthcare algorithm accuracy using survey and interview data to understand how online medical tools affect user decision-making.

Highlights:
· Analyzed 500+ data points.

· Conducted 50+ interviews and surveys.

· Wrote a 5,000-word research paper.

26 KB

Optimized model size

6 FPS

Estimated throughput

INT8

Quantized inference

Real-Time Blur

Selective privacy masking

Results:

Architecture:

Verilog

Vivado

CNN Inference

Python

Zynq-7000

ARM Cortex-A9

AXI-Lite

INT8 Quantization

Gaussian Blur Masking

Technical Stack:

I designed the system architecture, developed the lightweight CNN pipeline, optimized the model for edge deployment, and implemented the privacy-masking logic for real-time video processing.

Because the system needed to run under hardware constraints, I optimized the model through hard-negative mining, pruning, and INT8 quantization. This reduced model size to 26 KB while maintaining real-time feasibility, reaching an estimated 6 FPS against a 5 FPS target.

My Role:

Optimization:

I built an embedded computer vision system that detects privacy-sensitive visual content before frames are transmitted or stored. The system uses ARM Cortex-A9 inference to identify regions of interest and a custom Verilog masking block to apply selective Gaussian blur in real time.

Overview:

Real-time privacy filtering on edge hardware using a hybrid ARM/FPGA architecture.

FPGA-Based Embedded Vision System

· Electromyography signal input system
· Control logic in C++
· 3D-printed prosthetic structure
· Motor/control integration
· Multiple working prototypes

· Improved mechanical robustness by 70%
· Improved responsiveness by 40%
· Presented at four conferences/fairs
· Earned national awards

What I built:

Results:

Emerging Prosthetics is a cost-conscious prosthetic arm project designed to make prosthetic technology more accessible for amputees and non-amputees alike. I built the system using EMG muscle-signal input, C++ control logic, embedded hardware, and 3D-printed mechanical components. The prosthetic was presented at four conferences and I won various awards, including a 1st place prize.

Overview:

An EMG-controlled prosthetic arm built with embedded systems, 3D printing,
and real-time signal processing

Emerging Prosthetics

Learn more about my professional and leadership experience

Create a free website with Framer, the website builder loved by startups, designers and agencies.