About

A boutique R&D studio.

Rock R&D designs and deploys AI systems that operate where the data is created: on drones, vessels, industrial machines, and remote sensors. We specialize in taking models from research papers to real hardware — optimized for latency, power, and reliability in environments where cloud connectivity is a luxury.

Principal-Led

Rock R&D is led by its founding engineer — a defence scientist and ML systems engineer with an MEng (University of Toronto) and a BMath in Statistics & Applied Mathematics (Waterloo). His career spans the full stack of edge intelligence: research-grade perception for national defence, production AI toolchains at AMD, and real-time autonomous perception at Huawei. Engagements are deliberately limited in number.

MEng — UofTBMath — WaterlooTransport Canada RPAASA×UW Datathon winner
Principal — grayscale portrait (optional)
FOUNDING ENGINEER
Track record

Where the experience comes from

Each entry is attributed to its employer. This work belongs to those organizations — it is presented as résumé fact, not as Rock R&D project work.

2024–present

Defence Scientist

DRDC — Dept. of National Defence

AI research for maritime acoustic surveillance: real-time classification, bearing estimation, self-supervised pretraining for sonar (45% representation-quality gain over random init). EOIR detection/tracking for unmanned surface vehicles.

ACOUSTICSSSL
2022–2024

Software Engineer

AMD

Shipped 3+ CV models into the commercialized RyzenAI SDK; C++/Python APIs. >50% latency reduction and +20% accuracy on next-gen AI devices.

EDGEC++
2021–2022

ML Research Engineer

Huawei Canada

Custom DL operators under hardware constraints (up to 500% efficiency gains); BEV/perception model optimization to real-time (−60% latency); edge AI platform in Python/ROS.

PERCEPTIONROS
2020–2021

Graduate Researcher

University of Toronto

Infrastructure renewal-planning framework (Markov processes, Monte Carlo) minimizing lifecycle maintenance cost.

OPTIMIZATION
2020–2021

ML Research & AI Developer

CAAIE · Intellense / IBM

Bioinformatics deep-learning review; deployed license-plate recognition into a live surveillance pipeline.

VISION

Have a sensor, a site, or a fleet that should be smarter?

We take on a limited number of engagements per quarter. If your problem involves edge constraints, unusual sensors, or no labeled data — that's our specialty.

Book a technical consultationSend us your problem