About Me

I build real-time C++/CUDA systems where latency, memory, and correctness decide whether it works. Sole developer of SigTekX, a GPU-accelerated signal-processing engine I took from a one-line request to a packaged, tested, cloud-deployable framework. What I care about is the layer underneath: how code actually behaves on the machine.

Education

University of Phoenix

Bachelor of Science in Computer Science
Feb 2022 — May 2026

GPA: 3.93/4.00

Algorithmic Theory Software Architecture Calculus I & II Linear Algebra Discrete Math Computer Organization Operating Systems

Technical Skills

Languages

C++ (17) CUDA C/C++ Python 3 Bash PowerShell

Core Tech

CMake Docker Git PyBind11 GitHub Actions FastAPI Pydantic MLFlow Streamlit

Concepts

HPC (CPU/GPU) Real-time Systems Lock-Free Concurrency Memory-Bound Optimization Latency Hiding Signal Processing Zero-Copy IEEE 754

Tools

NVIDIA GPUs cuFFT Nsight Systems Nsight Compute Windows (MSVC) Linux (Ubuntu)

AWS

EC2 (Spot) ECR S3 CloudWatch IAM

Experience

Research Software Engineer (Volunteer)

SEAL Laboratory, University of Washington | Feb 2025 — Feb 2026
  • Sole Developer (SigTekX): Authored 95% of a 440-commit C++17/CUDA framework, scoping it from a one-line request ("enable CUDA") through architecture, optimization, packaging, and cloud deployment.
  • Architecture Decision: Prototyped GPU offload in Python (CuPy), measured it slower than the CPU baseline for real-time work, and pivoted to native C++17/CUDA — learning C++ from scratch to do it.
  • Lock-Free Concurrency: Implemented a zero-copy, thread-safe ring buffer using atomic acquire/release semantics to decouple acquisition from processing without mutex overhead.
  • GPU Acceleration: Built CUDA spectral pipelines (FFT/STFT) on cuFFT with multi-stream latency hiding and pinned-memory transfers, targeting a memory-bound CPU bottleneck; validated against recorded datasets on the target workstation GPU.
  • Numerical Correctness: Enforced IEEE 754 compliance using FMA operations, verified against NumPy reference implementations, with GPU clock locking to stabilize benchmark measurements.
  • Distribution: Packaged the compiled extension as a PyPI-published binary wheel via scikit-build-core, CMake, and PyBind11 — targeting Windows/MSVC first, then porting to Linux.
  • Cloud Orchestration: Architected a reproducible AWS GPU benchmarking pipeline (Docker → ECR → EC2, scoped IAM, S3, CloudWatch), then replaced the manual multi-script workflow with a FastAPI service that provisions, runs, retrieves, and tears down in a single request.
  • Test Engineering: Maintained a 422-test suite (pytest and GoogleTest) behind a coverage gate, with GitHub Actions CI running containerized lint and tests on cached Docker builds.

Interests

I'm most drawn to AI/ML infrastructure, LEO and satellite systems, and real-time and embedded work — different domains, but the same fundamentals underneath. Longer term I'm planning a master's in applied mathematics, because the math under these systems is what I want to get genuinely good at. Outside of that: PC building and gaming, snowboarding, mountain biking, hiking/nature, and motorsports.

Let's build something fast.

Email Me View Resume