About
Product management and engineering for geospatial and Earth observation data.
I manage and build geospatial data products. That means sitting with the people who will use a product, deciding what it has to do, and often writing a good part of the pipeline myself. Most of the work has been satellite and aerial imagery, wildfire risk and maps.
What I look for is whether the people a product was built for keep using it. The data-governance application I shipped for PG&E in ninety days is still in production after my contract ended. At YuzuData we tested two markets, chose real-estate site selection, and signed our first paying customers there; they changed their own site-evaluation process around what we built, and the detection pipeline is open source now as detr-geo. At Near I managed fifteen engineers and analysts and shipped six data products on a platform handling 1.7 billion mobility events a month.
The open-source work is product research I do on my own time. Mapping burn severity from Planet Tanager-1’s hyperspectral bands meant doing by hand everything analysis-ready data is supposed to automate, which is the quickest way I know to find what a data product is missing. I studied geography at San Francisco State and GIS at the University of Denver.
Where I’ve worked
Pacific Gas & Electric
Product manager, contract
I came back to PG&E to build an application for deciding which customer records can be disposed of, and for showing that each decision holds up. I wrote the requirements with legal, compliance and engineering, built it on AWS with one QA engineer alongside me, and made sure a person confirmed every step the model proposed.
–
Excella
Senior data consultant, backend and data infrastructure
On a public-sector engagement, in a Public Trust position, I designed and deployed a platform for detecting healthcare fraud, waste and abuse, built as a data-engineering product. I built its event-driven infrastructure on AWS serverless, the anomaly detection that runs as part of each day's pipeline, and the monitoring that tracks data integrity at every stage of the flow and raises an alert when something looks wrong.
–
Zoox
Cartographer, contract
I drew the ground truth for Zoox's map-generation model across San Francisco: lane lines, curbs, crosswalks, stop signs, traffic lights and drivable area, traced from the fleet's lidar, camera and imagery. The labels trained the model, and on the next pass I reviewed what it had built for errors and missed features. I also wrote Python checks against open datasets that caught my own mistakes and stale open data alike.
–
YuzuData
Co-founder
We built a site-selection platform for real-estate clients out of parcel, census and multifamily-market data, and a pipeline that found parked cars in satellite imagery. I ran customer discovery and designed the backend that fed the models and served their results.
–
Near
Manager, analytics and data engineering
I joined as a senior analyst building spatial pipelines for human mobility data and was promoted to run the team behind them. Two team leads reported to me, each with a team of their own, and much of the job was helping them grow as managers while we redesigned the pipelines and retired a decade of legacy code.
–
Pacific Gas & Electric
Spatial data analyst
I joined drone imagery, sensor telemetry, asset records and environmental layers in PostGIS and on AWS for PG&E's wildfire-risk and gas-corrosion models. During Public Safety Power Shutoffs I worked in the emergency response center, running the spatial analysis that found the switching path to keep as many customers' power on as possible.