Brent A. Thorne
Research Engineer • Embedded Systems Architect • Scientific Computing Specialist
Scientific DevOps • Real-Time Video + Sensor Systems • HPC Optimization • Applied AI
🌐 About
I build intelligent, resilient platforms at the intersection of embedded systems, high-performance computing, and scientific modeling. My work spans:
- Embedded Linux drivers and kernel integration (I2C, SPI, UART, CAN)
- Real-time sensor + video acquisition pipelines (GStreamer, V4L2, FFmpeg)
- AI/ML model deployment in constrained environments (PyTorch, CUDA)
- Scientific DevOps and reproducibility via CI/CD, Docker, and GitHub Actions
- HPC acceleration with SIMD (AVX2), OpenMP, MPI, and custom CUDA kernels
I specialize in projects that require low-level insight and high-level systems thinking.
📚 Education
- M.S. in Molecular Science and Software Engineering — University of California, Berkeley (2026)
- Certificate in Applied Data Science — Massachusetts Institute of Technology (2023)
- Open University at College of Science and Engineering – San Francisco State University (1999-2021)
- B.S. in Electronics Engineering Technology — Hamilton Technical College (1993)
💼 Selected Experience
- Freelance Research Engineer – Embedded Linux, radar systems, imaging platforms (2018–Present)
- Senior HW/SW R&D Engineer, Grid Net – Metering modem platforms, kernel/driver development (2013–2018)
- Systems & Integration Engineer, OpenTV – Embedded Linux, broadband media platforms, protocol stacks (2004–2011)
🧠 Projects
Kernel Driver for Spectral Radar Decomposition
Developed a Linux kernel-space driver for an embedded radar module. Streamed high-rate ADC data through ring-buffer DMA into real-time user-space FFT engines for spectral profiling and analysis.
→ Enabled low-latency RF signal capture and decomposition for radar imaging and diagnostics
Tools: C, Kernel Modules, DMA, mmap, FFTW, custom ring buffers
[repo coming soon]
Sphere Point Distribution for Agent Initialization
Relaxed uniform sampling of spherical domains using Fibonacci, geodesic, and Lissajous curves with CUDA-accelerated force modeling
→ Optimized for agent training and volumetric analysis
Tools: CUDA, Python, AVX2, Matplotlib
[repo coming soon]
Multi-Camera Stream Pipeline Using V4L2 + RTSP
Real-time synchronization across CSI/USB cameras with low-latency streaming via MediaMTX
→ Supports embedded vision applications on constrained compute
Tools: V4L2, FFmpeg, MediaMTX, GStreamer
[repo coming soon]
Disaster Image Severity Classification with CNNs
Fine-tuned convolutional models to classify damage severity from disaster-scene imagery using public crisis datasets
https://github.com/fractalclockwork/portfolio→ Explored multimodal cues for humanitarian response modeling
Tools: PyTorch, TensorFlow, NumPy, OpenCV
repo
🔧 Tools & Specialties
- Languages: Python, C/C++, CUDA, MATLAB, Rust (exploratory)
- Frameworks: TensorFlow, PyTorch, OpenCV, SciPy stack
- Embedded: Yocto, Buildroot, V4L2, DMA, Device Trees
- HPC: OpenMP, MPI, AVX2, CUDA kernels
- DevOps: Docker, GitHub Actions, CI/CD workflows
- Visualization: Matplotlib, VTK, Mayavi
📫 Contact
“Simple systems can grow complex behavior. Complex systems require careful simplicity.”