Machine Learning Engineer at ITPS Canada · M.E.Sc. Western University
I build ML systems that interpret human behaviour in high-stakes environments — from detecting pilot stress through physiological signals to designing sensor fusion pipelines for helmet-mounted displays. My work sits at the intersection of signal processing, computer vision, and applied machine learning, with a focus on real-time, safety-critical applications.
Currently building ML models for pilot stress detection from multimodal physiological signals at ITPS Canada.
Modeling Driver State and Takeover Behavior in Conditional Automation Using Physiological Signals
IEEE Open Journal of Intelligent Transportation Systems, Vol. 7 · 2026
At the Heart of Intersections: Analyzing Their Influence on Driver Heart Behaviour
Data Science for Transportation, Springer Nature · 2025
Navigating the Handover: A Review of Takeover Requests in Level 3 Autonomous Vehicles
IEEE Open Journal of Vehicular Technology · August 2024
AiX Lab Autonomous Driving Simulation Environment
Multi-modal simulation environment for autonomous driving research with synchronized data collection (video, physiological signals, vehicle dynamics) and integrated CV and signal processing pipelines for ML model training and validation.
Reinforcement Learning for Autonomous Driving
Deep RL agents (DQN, PPO) for autonomous navigation using CNN-based visual perception, trained on complex multi-agent driving scenarios in CARLA with varying weather and traffic conditions.
Helmet-Mounted Display (HMD) VR System for Pilots
A VR helmet-mounted display system for pilots with real-time information overlay using computer vision and sensor fusion. Processes multi-modal flight data (IMU, GPS, attitude sensors) through a Godot-based 3D rendering pipeline with quaternion-based orientation tracking.
lmnh — Local Speech-to-Text for macOS
Open-source menu bar app for macOS providing local speech-to-text and text-to-speech using Whisper — a privacy-first alternative to cloud-based dictation tools like Wispr and Speechify.
FoodSnap — Nutritional Analysis from Photos
ML application that estimates calories and macronutrients from a single meal photo using image segmentation, monocular depth estimation, and the USDA food database. Capstone project at Western.
RL Prisoner's Dilemma
Exploring how two RL agents behave when faced with the prisoner's dilemma — investigating emergent cooperation and defection strategies through repeated gameplay.
Micrograd from Scratch
A from-scratch implementation of an autograd engine and neural network library, building backpropagation and gradient descent from first principles.
Western AI Club — Waste Classification App
Led a team of five to build and deploy an iOS image classification app for waste categorization. Applied transfer learning to achieve 98% accuracy across waste types. Managed milestones and delivery as Project Manager.
ITPS Canada · London, Ontario
ITPS Canada · London, Ontario
Western University · London, Ontario
National Research Council Canada · London, Ontario
IBM · Markham, Ontario
Western University · London, Ontario
ECG signal generated with NeuroKit2 (ECGSYN model), processed through their signal processing pipeline — R-peak detection, PQRST delineation, and heart rate estimation. This is the kind of physiological signal analysis I did during my Master's research.
Signal: NeuroKit2 ECGSYN · 250 Hz · McSharry et al. (2003)
Helmet-mounted display system built for test pilots at ITPS Canada. Plays back real flight data through a Godot-rendered HMD with synchronized cockpit video, maps, and telemetry charts.
A 2D platformer built in Unity with C#. Features a boss fight, gem-collecting double-bounce mechanic, and a hidden door. Controls: W/Up to jump, S/Down to crouch, A/D to move.