Software developer

Hi, I'm a Honours Mathematics & Comp. Sci. U2 student at McGill University. I'm passionate about mathematics and love to learn how to apply abstract mathematical concepts into real-life solutions, especially in computer science, AI, and machine learning. I like to build projects from the ground up, showcasing interesting mathematical concepts or algorithms through applications, visual effects, or simulations.

I'm also a passionate guitar player, mainly playing instrumental/experimental rock such as Steve Vai or Joe Satriani 🎸.

Built a symbolic algebra engine from scratch to explore core algorithms behind symbolic manipulation.

• Core Features: Implements expression tree traversal for symbolic simplification, expansion, factoring, and differentiation.
• Tech: C# (.NET).
• Focus: Mathematical precision and modularity for complex algebraic rules.

A comprehensive progression from foundational ML algorithms implemented from first principles to modern deep learning architectures designed for COMP 551.

• Core Topics: Multivariable Regression, Logistic Classification (SGD with L1/L2 Regularization), Image Classification (Siamese CNNs/MLPs), and Sequence Modeling (LSTMs & Transformers).
• Tech Stack: Python, PyTorch, Hugging Face transformers, Scikit-learn, NumPy, Pandas, Jupyter, LaTeX.
• Key Highlights:
  • Tabular Data: Engineered a custom multivariable regression model for bike-sharing demand, achieving a 17.7% MSE reduction through advanced polynomial feature interaction.
  • Computer Vision: Designed a custom PyTorch Siamese CNN and Geometric MLP for an "Odd-One-Out" Kaggle competition, strictly adhering to a <25k trainable parameter limit.
  • NLP / Sequence Modeling: Evaluated text classification strategies by comparing the training dynamics of a from-scratch word-level LSTM against a fine-tuned DistilBERT transformer (achieving 94.4% accuracy).

Industrial R&D project developed during my internship at Carbotech.

• AI & CV: Leveraged Machine Learning classification models and Computer Vision algorithms for real-time species detection on lumber boards.
• Implementation: Engineered the logic layer in C++ and managed high-level UI and server communication in C#.
• Scope: Includes camera calibration modules and integration with industrial machinery and SQL databases.

• G-PST: Graph-Based Pathfinding 📍
  • Implementation of the A (A-Star)* search algorithm on a weighted graph (nodes and edges) to calculate optimal routes in an interactive 3D map.
• 3D Procedural Terrain Generation 🌳
  • Showcases the use of 2D and 3D Perlin Noise to create semi-realistic, customizable terrains and cloud formations.
• Metaballs 🧶
  • An artistic visualization of the mathematical concept of metaballs and 2D function illustrations.
• Pi and Bouncing Blocks 🥧
  • A simulation calculating the value of $\pi$ using the physics of perfectly elastic collisions between blocks of varying masses.

• 3rd BlackThornProd Game Jam
  • My submission: Just an Easy Dungeon

• Dean's Honour List (2024-2025) - McGill University.
• Faculty of Science Scholarship - Awarded for being in the top 5% of the Faculty.
• TEAM Mentor - Selected to mentor 500+ students in Intro to Computer Science (COMP 250).