Projects — ML + Software
Pipelines, evaluation, and product thinking in practice. Building artifacts informed by real-world constraints.
After Synkron, I wanted to go beyond using machine learning as a black box. I started building systems myself — from data pipelines and model training to full-stack apps that put AI in users' hands. My focus: data, machine learning, and software that solve real problems.
These projects are where theory meets practice. Each one pushes on a different part of the stack: data wrangling and feature engineering, model evaluation and iteration, or end-to-end product design. The goal is not perfection — it's learning how systems fail, scale, and ship.
What I built
Across data, ML, and software, the work falls into three buckets:
- Data & Machine Learning
House-Price-ML (v1 baseline and v2 feature engineering) — full ML pipelines from EDA to scikit-learn models. Pendulum Data-Driven Control — reinforcement learning for control theory. ML Hackathon project (CatBoost for overqualification prediction). - Full-stack / Software
GymWhisper — voice-powered workout tracking (React, Gemini API, speech recognition). TelusGuardAI / Network Impact Analyzer — multi-agent AI for geospatial network outage analysis (React, Flask, Leaflet). - Systems & Coursework
Personal website, portfolio projects, and research on AI in Autonomous Vehicles. Smaller experiments in C, Linux, and CLI tooling.
Tools
Python (scikit-learn, Pandas, NumPy, PyTorch) for data and ML. React, Node.js, and Flask for web. Git, Linux, and structured workflows for shipping. The stack evolves with each project — the constant is thinking in terms of pipelines, evaluation metrics, and user value.
Proof / demos
See the full project catalog with live demos, reports, and write-ups: Projects. Highlights include GymWhisper (demo + report), House-Price-ML (report), and TelusGuardAI (demo + report).