Experience & Education

Undergraduate Research Fellow

Rutgers University | Apr 2025 – Present

Competitive fellowship in the Physical AI Lab focusing on AI & robotics.

Key Responsibilities:

  • Interdisciplinary AI & robotics research
  • Prototype novel perception systems
  • Publish results & present to faculty

AI/ML & Data Engineering Intern

Atyeti Inc. | May 2024 – Sep 2024

Developed ML models & automated data processing.

Key Achievements:

  • Vision Transformer digit model (98 % accuracy)
  • Automated cleaning (-18 hrs per project)
  • Data validation (97 % analysis accuracy)

Engineering Lead & Software Engineer

Rutgers VEXU Robotics | Sep 2024 – Present

Lead engineering & software for competitive robotics team.

Key Responsibilities:

  • Autonomous & driver code (C++)
  • Sensor fusion & motion planning
  • Mechanical design leadership

Full-Stack Web Dev Intern

Team Tranquil Inc. | Jul 2023 – Dec 2023

Developed full-stack web solutions for clients.

Key Achievements:

  • Restaurant website w/ SQL backend
  • Image management system (+50 % productivity)
  • Optimized SQL queries (+33 % speed)

B.S. Computer Science & Data Science

Rutgers University | 2024 – 2027 (Expected)

GPA 4.0 | Specializing in AI/ML & Robotics

Activities & Societies:

  • Physical AI Research
  • VEXU Robotics (Lead)
  • Competitive Programming
  • RU Autonomous
  • USACS

My Projects

HoloBrain

HoloBrain

ML/CV

AI-powered 3D MRI hologram with real-time hand-gesture controls for faster, more accurate brain-scan diagnosis.

PyTorch OpenCV  MediaPipe Unity 3D

Detailed Description

HoloBrain couples a 97 %-accurate CNN classifier with a 3-D volumetric renderer to turn conventional MRI stacks into an interactive hologram. Doctors can rotate, zoom, and slice the brain model using 6 natural hand gestures detected by OpenCV + MediaPipe, leaving the keyboard and mouse behind. A lightweight Python-to-Unity socket keeps classification overlays and gesture input in sync at 30 fps, cutting review time by 40 % in user testing.

CNN (97 %) Gesture Control Real-time Sockets
Handwriting Recognition

Handwriting Recognition

ML

EMNIST letter-classifier achieving 92 % accuracy using a 6-layer CNN with data augmentation and hyper-tuning.

PyTorch CNN Python

Detailed Description

Trained a custom convolutional network on 130 k EMNIST characters with on-the-fly rotation, shear, and elastic transforms. Implemented learning-rate warm-up and cyclical scheduling to squeeze out 4 % extra accuracy over baseline. Results are served through a Flask API that returns top-3 predictions in under 20 ms.

Data Augmentation Flask API 92 % Top-1
Restaurant Website

Indian Cuisine Restaurant

Web Dev

Responsive ASP.NET webapp featuring live menu, geocoded locations, and an admin portal backed by SQL Server.

ASP.NET Core Bootstrap SQL Server

Detailed Description

Multi-page site lets diners browse a photo-rich menu, place pickup orders, and view promos nearest to them via Google Maps embedding. The owner dashboard supports CRUD for dishes, photos, and coupons, all secured with ASP.NET Identity. Optimised SQL indices dropped page-load P95 from 1.9 s to 1.2 s.

E-Commerce Role-Based Auth Performance Tuning
Digit Recognition

Digit Recognition

ML

Side-by-side CNN vs. Vision Transformer for MNIST digits, topping out at 98 % accuracy with 30 % fewer params.

PyTorch CNN vs ViT TensorBoard

Detailed Description

Re-implemented LeNet-5 and a compact DeiT-style transformer to probe where attention outperforms convolutions. Benchmarked training curves, FLOPs, and Grad-CAM saliency maps; published findings showing transformers closed the gap after 8 epochs while being 22 % faster at inference on GPU.

Model Benchmarking Grad-CAM 98 % Acc
Robotics

VEX Robotics

Robotics

C++ control stack for Rutgers VEXU robots: autonomous navigation, PID, and driver assist for top-5 regional finish.

C++ Sensor Fusion PID Loops

Detailed Description

Designed modular task scheduler around PROS RTOS, fusing IMU, encoders, and vision to achieve ±1° orientation repeatability. Dynamic motion-planning cut autonomous cycle time by 18 %. Driver-assist modes add auto-aim and intake sequencing triggered by controller macros.

Autonomous Ops Real-Time OS Competition Bots
Data Processing

Data Processing

Data

Excel-Python ETL toolkit that slashes manual cleaning by 18 hours per dataset via smart type-casting and validation.

Python Pandas Excel Interop

Detailed Description

Automated detection of mixed dtype columns, date normalisation, and cross-sheet referential checks push dataset integrity to 97 %. Generates reproducible cleaning logs and summary dashboards for business analysts, delivered as a one-click macro-enabled workbook.

ETL Data Quality 97 % Automation
VEXAI Strategy Planner

VEXAI Strategy Planner

AI/Robotics

Pre-match AI that explores 10 k+ action trees in 1 s to output the highest-value autonomous paths for VEXU.

Python MCTS Reinforcement Learning

Detailed Description

Generates a randomised replica of the High Stakes field, then simulates pick-and-place actions with physics-aware cost models. A Monte-Carlo Tree Search prunes low-value branches while a Q-learning critic refines move ordering between iterations, converging to strategies that score 15 % higher than human baselines. Outputs JSON playbooks that sync to the competition robot over Wi-Fi.

Procedural Field Gen Q-Learning Critic JSON Playbooks
View All Projects

My Skills

Languages

  • Python
  • Java
  • C#
  • C / C++
  • JavaScript
  • HTML / CSS
  • SQL

Developer Tools

  • Git / GitHub
  • Visual Studio & VS Code
  • IntelliJ IDEA
  • PyCharm
  • Unity
  • Android Studio
  • Docker
  • Microsoft Azure

Frameworks & Libraries

  • PyTorch
  • OpenCV
  • MediaPipe
  • React JS
  • Bootstrap
  • ASP.NET Core
  • jQuery
  • Tailwind CSS
  • Socket IO
  • NumPy / Pandas

Interests

  • Artificial Intelligence
  • Machine Learning
  • Computer Vision
  • Data Science
  • Robotics & Autonomous Systems
  • Computer Science
  • 3D Visualization & XR
  • Cloud Computing

Technologies & Tools

Python
PyTorch
OpenCV
Java
JavaScript
React
SQL
Git