I'm a multidisciplinary engineer focused on models: how to build them, interrogate them, and turn them into systems people can act on.
My background spans electronics engineering and cognitive science, connected by software, statistics, and machine learning. It taught me to move between mechanisms and behavior, and to question what a model represents, where it fails, and whether its output is reliable enough to support a decision.
At Microsoft, I work on making agentic qualitative analysis of unstructured data more reliable and controllable: workflows that can be rerun, outputs that remain traceable to their sources, and analyses that can be inspected rather than merely accepted.
Azure capacity planning 路 Prescriptive models and simulation
Built models and simulation engines for global cloud capacity, balancing inventory and GPU/data-center rollout lead times against forecast demand.
PACCAR 路 Reliability and explainability
Built explainable models for predicting mechanical failures. In 2019, before language models became synonymous with AI, I also built small, domain-specific language models that translated field-engineering shorthand into normalized failure modes and connected field evidence back into the FMEA process. Explainability was part of the model, not a reporting layer added later.
Innoplexus / Partex 路 Knowledge representation
Built ontologies and information architecture for pharmaceutical research, including Asset42 and Meta-D3.
The domains have changed, but the problem keeps rhyming: turn messy evidence into models people can act on without losing the path from evidence to answer.
- Built a wireless split keyboard, its firmware, and a cross-platform shortcut overlay.
- Use simulation and rendering to design furniture and unusual shapes from that surprisingly scarce material: wood.