Teodor Ilie
I build vehicles that drive themselves.
On the Autonomy Team at MacLean Engineering, building self-driving underground mining vehicles.

- Built
- Gym-Khana deep RL simulator
- Founder
- Ingenuity Labs Racing
- Based
- Ontario, Canada
From simulation to the real world
My Master's research at Queen's, with Ingenuity Labs and the QUARRG lab, asked whether deep reinforcement learning can control a car at the limit of grip. To find out I built Gym-Khana, a parallelized simulator for training policies to race and drift, then took those policies off the screen and onto a physical 1/10-scale car.
Alongside the research I founded Ingenuity Labs Racing, leading PhD, post-doc and MSc students through building a complete autonomous racing stack, from SLAM and planning to control. Four years as a full-stack developer at BMO taught me to write software other people depend on. Now I bring all of it to real machines on the Autonomy Team at MacLean Engineering.
Outside of work I train in kickboxing, play guitar, and design LEGO Technic vehicles from the ground up.
Interests
- Reinforcement Learning
- GenAI
- Autonomous Control
- Event Camera
- Computer Vision
- Deep Learning
- Automated Planning
- Explainable AI
- Cybersecurity
- FinTech
Education
- M.Sc., Computer Science (AI Spec.)Queen's University · 2024 – 2026
- B.Sc. (Hons), Computer ScienceQueen's University · 2018 – 2022
Research and industry
Oct 2026 – Now
Designer Level II, Autonomy Team
MacLean EngineeringBuilding autonomous systems for underground mining vehicles.
Jul 2025 – 2026
Founder & Team Lead
Ingenuity Labs RacingFounded and led Queen's F1TENTH / RoboRacer autonomous racing team of PhD, post-doc and MSc students, building ROS2 SLAM, planning and control stacks. Hosted Canada's first RoboRacer event.
See the projectSep 2024 – 2026
Graduate Research Engineer
Ingenuity Labs, Queen's UniversityDeveloped deep RL controllers for autonomous vehicle drifting and loss-of-traction recovery, trained in a custom parallelized Gym simulator and deployed on a physical 1/10-scale car.
See the projectSep 2022 – 2026
Full-Stack Software Developer
BMO Financial GroupBuilt SKOPE, an internal safekeeping app with Spring REST APIs and an Angular front end, in an Agile team practising TDD.
May 2020 – Aug 2021
Business Analyst (Summer Internships)
BMO Financial GroupDesigned an automated Power BI reporting tool later adopted bank-wide, and coordinated month-end systems monitoring across teams.
Selected work
Autonomous racing, reinforcement learning for robot navigation, probabilistic inference, and neural networks built from scratch.
CNN Handwriting Classifier
Designed and implemented a Convolutional Neural Network model in Python from scratch to classify handwritten digits, using the MNIST dataset.
- Python
- Convolutional Neural Network
AI Pacman Game
Implemented probabilistic inference algorithms to enable Pacman to hunt invisible ghosts based on noisy distance readings.
- Inference
- Python
- AI
Monocular Visual Odometry in OpenCV
Implemented SIFT feature detection and brute-force matching to accomplish Monocular Visual Odometry (MVO)
- OpenCV
- MVO
- Carla
- AdverCity
OpenGL Scene Generation
Designed a 3D scene with ray tracing, reflection, refraction, and surface rendering in OpenGL using C++
- Computer Graphics
- Ray Tracing
- OpenGL
LEGO Technic
Over a decade of designing working vehicles from scratch: custom gearboxes, transmissions, suspensions and axles, each engineered from original blueprints. The same instinct drives my software work: building a great system isn't so different from engineering a really capable truck.

- Heavy-duty gearbox
- 6x6
- +2

- Pneumatics
- 4x4
- +3

- Dakar
- MAN
- +2

- RAM
- Pickup truck
- +2

- BuWizz
- Baja truck
- +2

- Half-axle
- Torsion bar
- +2

- Rally
- Audi
- +1

- Mercedes
- Pendular Suspension
- +1

- Tatra
- Skid-steering
- +1

- Ural
- Live axles
- +1

- Mercedes
- Trial Truck
- +2
Builders who inspire me
Let's talk
Open to research collaborations, engineering roles, and LEGO Technic chats.
Elsewhere
Message sent. I'll get back to you soon.
Something went wrong. Please try again later.
Sent securely and never shared with third parties.