01
From records to a score per cell
Schematic with synthetic values. Real outputs belong to the clients.
02
What was built
- A commercial-behaviour database integrating 300K+ Census, ACS, point-of-interest and mobility records, keyed to a common spatial grid so that any indicator can be compared across places.
- A 10+ indicator evaluation framework — demographics, accessibility, competition, cash-flow proxies — for site selection and growth, with each indicator defined and sourced so that a client can argue with it.
- Research automation: Python and language-model workflows that took data processing and first-draft client reporting off the analysts' desks.
- Decision research on market entry, site selection and growth for 20+ Series A–C AI and consumer-tech companies.
Why it belongs with the AI work
It is the same discipline in a different medium: define what a score means before computing it, keep every number traceable to a public source, and separate what the data can say from what the client must decide. The architecture training behind it — reading a site, its constraints and the behaviour it will hold — is where that habit comes from.
Records
300K+ Census, ACS, POI, mobility
Framework
10+ indicators across four families
Clients
20+ Series A–C AI and consumer-tech companies