Projects

Below is a list of my projects across machine learning, application development, and electrical engineering. Please click on each project to learn more!

Machine Learning

LLMs for Automatic Hardware Testbench Generation

A machine learning project exploring how large language models can automate the generation of hardware testbenches.

This project implements an automated Verilog testbench generation system using Large Language Models (LLMs). It fine-tunes open-source LLMs to generate functional testbenches for hardware designs described in Verilog HDL.

Features

• Automatic testbench generation from Verilog DUT code
• Fine-tuning pipeline using LoRA/QLoRA for efficiency
• Comprehensive evaluation metrics (compilation, simulation, coverage)
• Support for multiple datasets (AutoBench, MG-Verilog, HDLBits)
• Experiment tracking with Weights & Biases

Technical Approach

Use transformer‑based models to generate testbench code from hardware specifications and integrate with existing HDL verification frameworks.

Adaptive Crowd Management AI

An AI system designed to dynamically manage crowd flow and safety using real‑time data.

An AI-based system designed to model, simulate, and predict crowd behavior in emergency situations within virtual environments. The system leverages multi-agent simulations, adaptive pathfinding, emotion analysis, and human-in-the-loop interventions to optimize emergency responses.

Features

• Multi-Agent Simulation Framework
• Adaptive Pathfinding Algorithms
• Emotion and Panic Level Analysis Using Neural Networks
• Interactive Human-in-the-Loop Dashboard
• Data Visualization and Post-Simulation Analytics

Technical Approach

Combine computer vision with reinforcement learning to manage crowd flow and safety, using simulation environments for training.

Applications

Clinic Management System

A web application to streamline clinic operations, scheduling, and patient records.

The semester project for CS5010 Fall 2024 @ Northeastern University is a fully implemented clinic management system which uses the Model-View-Controller (MVC) design pattern which organizes an application's logic into three components, the model, the view, and the controller. The goal of the three seperate milestones is to implement each portion of the MVC architecture, building upon previous milestone work. Teaching students about writing well thought out code and documentation, along with giving students experience with working on a large scale project.

Features

• Register a new patient
• Register a new clinical staff member
• Display information about a specified patient
• Display information about a specified room
• Display information about all rooms
• Assign a patient to a room
• Assign a clinical staff member to a patient
• Send a patient home
• Deactivate a staff member
• List active staff with current patients
• List inactive patients
• Patient visit summary for the past year
• Unassign clinical staff from a patient
• List all clinical staff and patient counts
• Add a new room
• Exit

Technical Approach

Built with a MVC design pattern.

Personal Homepage

This project—the personal homepage itself—built with vanilla HTML, CSS, and ES6 modules.

A showcase of my projects, skills and resume. Includes dark/light theme toggle and a tag filter component.

Key Objectives

• Demonstrate proficiency with vanilla web technologies.
• Create a clean, responsive personal site.

Technical Approach

Implemented with modular ES6 JavaScript, CSS Grid/Flexbox for layout, and no external frameworks. Deployed as a static site.

Electrical

Robotic Grasping Mechanism Capstone

A hardware and software project developing a robotic hand mechanism with precise grasping capability.

A senior capstone exploring electrical design and control of an robotic gripper for space debris removal applications.

Key Objectives

• Design a dexterous robotic hand capable of delicate grasping.
• Implement motion control and feedback for precision.

Technical Approach

Combines mechanical CAD design with motor control circuitry and microcontroller programming to achieve responsive hand movements.

NASA Moog Gimbal

An engineering project involving the design and control of a gimbal system for aerospace applications.

This project focuses on the design and control of a two‑axis gimbal system used in aerospace instrumentation.

Key Objectives

• Develop a stable, responsive gimbal algorithm for precise orientation control.
• Integrate control algorithms for smooth, accurate tracking.

Technical Approach

Utilizes servo motors and sensor feedback to control gimbal movement, paired with kinematic modelling and PID control techniques.