House hunting in Chicago’s Lake County suburbs meant checking flood maps, school ratings, traffic noise, and assessor records across half a dozen disconnected sites for every address on my list. Some of this is aggregated on sites like Redfin, but it stays too high level to act on and never gives enough context to actually judge a specific address. I designed the product as I would a finance model: define the inputs that actually matter, decide what "good" and "risk" mean for each one, and never assert a number the underlying source doesn’t support.

I worked with a GenAI coding agent to turn that spec into a shipped tool, sourcing and wiring up First Street, FEMA, GreatSchools, NCES, the DOT noise map, and county assessor data, then iterating on the read: a check only shows good, risk, or unknown, and only when real data backs it up.

Prepping for an interview usually means the same scattered research every time: reread the job posting, dig up recent company news, guess at culture and financials, and try to find something useful on the recruiter or hiring manager. It is a lot of low-value searching for a handful of things you actually need to say out loud in the room.

I worked with a GenAI coding agent to build a tool that compiles all three briefings live from the public web: talking points matched against your own resume for a specific job posting, a company briefing covering news, culture, and financials, and a contact briefing on the person you are meeting.