Hi, I'm Anthony Njenga. I'm an aspiring systems and performance engineer with a deep interest in computer architecture and GPU computing.
I'm passionate about writing low-level software that extracts maximum hardware efficiency — from cycle-accurate CPU microarchitecture simulations and hardware logic design in modern C++ to high-throughput CUDA kernels and deep learning systems from scratch.
I'm aspiring to work in Systems & Performance Engineering, High-Performance Computing (HPC), and GPU Kernel Engineering.
An intuitive breakdown of what GPUs are and how they differ from CPUs. Walks through core hierarchies and a CUDA vector addition kernel.
Implements and benchmarks CPU microarchitecture designs in modern C++. Spans single-cycle and multi-cycle execution through superscalar out-of-order pipelines with fetch, dispatch, Reorder Buffer, Issue Queue, Physical Register File, and branch recovery.
A modern C++20 hardware simulation library for modeling digital logic circuits from electrical primitives to complex architecture components. Models wires, buses, clocks, tri-state buffers, gates, combinational datapaths, and sequential storage elements.
Custom GPU kernel implementations tested on NVIDIA hardware (Tesla T4). Includes work-efficient parallel prefix scans (Blelloch & single-pass with decoupled look-back), tree-based reductions, and memory coalescing optimizations.
A lightweight, ground-up deep learning library with a dynamic computation graph and reverse-mode automatic differentiation. Includes core neural network modules, attention mechanisms, and model quantization strategies.
Implementations and benchmarking of classical and modern computer vision architectures: DeepLabV3, UNet, DenseNet-121, MobileNetV1 & V2, and SqueezeNet.