Nick Armstrong

• Studying computer engineering at the University of Waterloo
• Interested in embedded/firmware, tinyML, robotics, and high performance compute
• Interned at Onsemi and Ford Motors building performant embedded systems, and at Endex.ai building autonomous finance agents
• Always building something and down to chat!

Recent Posts

Selected Projects

Distributed CAN Network

STM32, Embedded C++, CAN, DMA

• Led firmware development for distributed STM32 system of plug-and-play nodes, using a shared CAN bus to synchronize behaviour with the drone state.
• Used the STM32 HAL to control DMA, interrupts, and hardware timers to communicate with board peripherals while minimizing overhead usage.
• Wrote modular, reusable, and highly maintainable embedded C++, making architectural decisions to decouple separate components.
• CAN LED network whitepaper

Low Power Semantic Visual SLAM

Rust, Assembly, SIMD, ESP32

• Handwritten image processing kernels in assembly, using SIMD and minimizing data dependencies to eliminate pipeline hazards and 4x throughput.
• Fine-tuned and quantized image embeddings model in TensorFlow, optimizing for high-throughput ESP-NN inference.
• Developed a PnP RANSAC pipeline for robust camera localization, rejecting dynamic outliers.

Robotic Path Planning Library

Trajectory Planning, Kalman Filtering, PID

• Use RProp numerical optimization to generate optimal Hermite spline trajectories.
• Developed Kalman-filter based odometry and IMU-based sensor fusion to adaptively react to disturbances.
• Deployed to a Linux device at FRC robotics competitions, enabling complex autonomous routes to score more points.

Closed Loop Stepper Motor ASIC

Verilog, TinyTapeout

• Built a FSM-based SPI peripheral, using best practices such as cascading flip-flops to avoid metastability issues.
• Implemented an in-hardware PI controller using accumulators and power-of-2 coefficients to eliminate unnecessary MAC units.
• Implemented motion planning math using a FSM, saving 4-5x area over the naive MACs and divide units.

Functional Programming Language Compiler

Rust, Git, GitHub

• Developed a compiler for a functional programming language using techniques including specialization, unboxing, and defunctionalization to maximize performance.
• Used the Rust type system to guarantee correctness of program transformations.