Projects
Work where physics, messy data, and operational decisions meet. Industry projects are described qualitatively; open-source projects link to code and live demos.
Applied ML for hardware reliability
Repair recommendation, failure diagnosis, and decision systems built on noisy operational data.
Hardware diagnostics and repair recommendations
Industry workProduction ML decision systems that fuse server telemetry, logs, and repair history into calibrated, component-level repair recommendations for AWS EC2.
- Python
- PyTorch
- SageMaker
- Time-series ML
- Calibration
Multimodal sensing for Just Walk Out
Industry workCapacitive and RFID sensing science for AWS Just Walk Out, turning multi-channel sensor time series into reliable per-item event decisions.
- Signal processing
- CUSUM
- Bayesian inference
- Transformers
- RFID
Physics-based engineering and simulation
Electromagnetics, phased arrays, and antenna design, modeled in code and driven by agents.
Agentic Phased Array Builder
An LLM agent that designs phased-array antennas end to end, driving electromagnetic solvers and array models through the Model Context Protocol.
- Python
- LLM agents
- MCP
- EdgeFEM
- Phased-array modeling
Phased Array Antenna Model
Live demoA vectorized Python library for phased-array antenna patterns (mutual coupling, phase quantization, beam steering), with a live interactive demo.
- Python
- NumPy
- SciPy
- Plotly
- Streamlit
Open-source tools for engineering decisions
Solvers, optimization packages, and feasibility studies for physical infrastructure.
EdgeFEM
An open-source C++20 finite-element electromagnetic solver for RF and mmWave: S-parameters and radiation patterns from 100 kHz to 110 GHz.
- C++20
- Python
- Eigen
- Gmsh
- pybind11
Orbital data center feasibility (spacedc-mdao)
Live demoA multidisciplinary design and optimization study of space-based data centers, accounting for power, thermal, network, and reliability constraints together.
- Python
- OpenMDAO
- pymoo
- SALib
- Streamlit
More on GitHub
Other packages and research code, available at github.com/jman4162.