TARDIS Prototype
I built this TARDIS-inspired AI chatbot as a lightweight, web-deployable conversational assistant intended for simple, natural conversations. It supports both text and audio input/output via browser APIs.
Beyond basic chat, the project implements a simulated long-term memory framework for maintaining conversational context, client-side cipher decoding tools (Caesar / ROT13) aligned with the TARDIS theme, and a secure serverless AI proxy to protect API credentials.
The system features a one-click landing experience suitable for non-technical users and is fully deployable as a static frontend with a serverless backend, requiring no local runtime for end users.
Repo Live
Portfolio Optimization Agent
This project is an end-to-end machine learning system that demonstrates how sequential decision-making models can combine time-series market signals with natural-language news sentiment to produce conservative, interpretable portfolio recommendations.
The system is designed to showcase model design and optimization, feature engineering, reinforcement learning, and production-oriented deployment, reflecting core skills expected of a Machine Learning Engineer.
Repo Live
AR Experience
This repository contains a lightweight web app that combines two playful computer-vision demos into one site. Both demos run entirely in the browser and require camera permission to function.
Cat Reaction is an interactive facial-expression detector that maps your face to popular TikTok cat reactions using MediaPipe Face Mesh, ported from the original Python MeowCV project.
Fruit Ninja is a simplified browser-based fruit-slicing game where falling fruit is overlaid on your webcam feed and sliced by swiping your index finger using MediaPipe Hands. The game lasts 60 seconds and updates the score in real time.
Repo Live
Attention Seeker
Attention Seeker is a full machine-learning pipeline designed to estimate a user’s attention level using physiological and motion data collected from wearable sensors.
The project computes a derived Attention Score using heart rate, HRV approximation, and wrist movement, as well as a modeled Outside Factors Score incorporating sleep and screen time.
It evaluates Pearson correlation between attention and outside factors and trains regression and classification models to predict cognitive load and attention lapses. The project is implemented entirely in Python using Jupyter Notebooks.
Repo
Heart Disease Risk Assessment
This project explores clinical and lifestyle risk factors associated with heart disease, emphasizing exploratory data analysis (EDA), reproducible reporting, and analytics-ready project structure.
The workflow mirrors real-world data analytics practices with a focus on data cleaning, visualization, and interpretation rather than black-box prediction.
Repo Live
Imaging Operations & Analytics
This showcase demonstrates familiarity with imaging informatics workflows including PACS concepts, HL7 interfaces, operational analytics, end-user support patterns, and security-minded system design.
The live demo runs on synthetic data, while the repository includes backend and integration scaffolding that reflects appropriate architecture boundaries for clinical systems.
Repo Live
Amazon Review Sentiment
This is an end-to-end natural language processing project that predicts sentiment and star ratings from product reviews, mirroring the workflow of an applied scientist working with large-scale e-commerce text data.
The repository is GitHub Pages-ready for interactive exploration and experiment-ready via Python scripts in the
src directory.
Fast Flow
Fast Flow is a lightweight, state-driven productivity dashboard where the UI adapts to user actions such as mode selection, widget toggles, layout density, and theme preferences.
It includes Study, Deep Work, and Chill modes; widgets like Pomodoro, Notes, Today’s Goals, and Quick Links; dynamic conditional rendering; persisted preferences via localStorage; and zero backend deployment.
Repo LiveSupply Chain Inventory API
A lightweight backend service demonstrating RESTful API design, data modeling, basic forecasting, and containerized deployment.
The system simulates inventory workflows including stock tracking, CRUD operations, and a forecasting module that predicts days until stockout using consumption rates. Built with FastAPI, SQLite/SQLAlchemy, Docker, and automated API documentation.
Repo
NFL Analysis
This project analyzes how NFL teams allocate salary cap resources to quarterbacks and running backs and how effectively those investments translate into team success.
The analysis spans the 2015–2025 seasons (excluding 2021 due to data unavailability) and uses Python, pandas, NumPy, and Matplotlib. The interactive site is styled using official NFL team colors.
Repo Live