Open-source privacy-preserving API and web app to safely and reliably interact with third-party LLMs such as ChatGPT and Cohere.

Python, FastAPI, TypeScript, React

Trained/optimized ML models (Logistic, Random Forest, and LGBM Classifiers) to sort Spotify tracks into genres. Correctly classified 90% of tracks, a 67% improvement from the baseline, and achieved 90% precision, recall, and f1 scores.

Python, Scikit-learn

Calculates and visualizes technical indicators with Pandas/NumPy including: moving averages, standard deviations, Bollinger bands, and RSI.

Python, Pandas, Matplotlib

Designed, normalized, and created a relational DB using Oracle SQL, Java Database Connectivity (JDBC), and Swing. Aggregates company/branch-wide information, and allows clients to register, search available rentals, and calculate estimated cost.

Java, SQL

Interactive Bokeh/Flask dashboard to visualize Canadian pandemic data by province. Calculates daily increase in COVID-19 cases by parsing JSON API data.

Python, Flask, Heroku

Stacks is a block-stacking game. Your goal is to stack the block as long as you can. If you miss, the block will get smaller and smaller until there's nothing left!

Python, Pygame

Employs multivariate regularized regression to predict car resale value, achieving an R² of 90.27% on test data.

Python, Scikit-learn

Notes and project work from Stanford's ML Course taught by Professor Andrew Ng on Coursera.

Python, NumPy

Solved various ML problems including: sentiment analysis, spam classification, churn/survival, and time series forecasting. Explored data pipelines, hyperparameter optimization, ensembles, transfer learning, and unsupervised learning.

Python, Scikit-learn

Finds the most efficient path for a taxi driver agent to pick-up and drop-off passengers using the Q-Learning (Reinforcement Learning) algorithm, achieving an average reward of 8.3 (world record is 9.7).

Python, OpenAI Gym

Solves 9x9 Sudoku puzzles using a backtracking depth-first search over a 2D array. Compiled for both Windows and Linux.

C++, MSVC, G++

Implementation of the classic "Space Invaders" game, created to explore object-oriented game development.

Python, Pygame

Predicts whether a borrower will default on their loan using K Nearest Neighbors (77.813% accuracy).

Python, Scikit-learn

Projects and coursework from CPSC 422 - AI II. Covers topics such as Reinforcement Learning, Belief Networks, Markov Networks, and Natural Language Processing.

Python, Jupyter

Solves a constraint satisfaction problem (CSP) using DFS with pruning. Optimized by a variable selection heuristic.

Python, Jupyter