Sell → Build → Scale

What this is

Most case studies describe what got built. This one is about what got decided — the real tradeoffs, rejected alternatives, and engineering judgment behind an AI data pipeline I sold solo, built solo, and scaled into the infrastructure other teams now run on. Every mechanism and decision is real. Every specific data point (place names, IDs, exact figures) is synthetic, reconstructed to match production shape without disclosing a confidential client relationship.

Eight chapters

• Boots on the Ground — why I shot ~400 storefronts myself before writing a line of code.
• An MVP That Delivers From Day One — what to build the human-review layer out of, before scale exists to justify anything expensive.
• Global-Scale Deployment and Operation — the hiring-to-payroll loop that lets a small admin team run 100+ workers without micromanaging.
• The Industry's Long Problem: When and Where to Update — finding new venues from public-record signals before they're searchable anywhere.
• A Precision Instrument, Not a Safety Net — why a verified contradiction between sources is the right signal for human review, not "low confidence."
• Scaling as a Data Partner, From One Pilot — selling a third and fourth capability into the same account, without a new sales cycle.
• Sharing the Tech Stack Across the Entire Org — what actually generalizes to other teams (the ops layer) and what doesn't (sourcing).
• How the Entire Org Now Runs Its Sales + Dev Cycle — the recurring role I hold on every new deal: sales sources it, I decide if it's real.

Open the interactive build log →

About Tae Hyun Koh

Tae Hyun Koh is an AI Product Lead and systems builder based in New York City. He specializes in enterprise AI products, HITL operations, and zero-to-one product development. He is currently open to Senior PM and AI Product Lead roles.

kohtae.com · taehyunkoh0607@gmail.com · LinkedIn