A forward deployed engineer in New York who can be in the room

Because the exceptions that break an AI workflow are rarely written down. You find them by sitting next to the person who does the work.

Take invoice approval. On paper, invoices arrive by email and get approved in the ERP. In practice there's a shared spreadsheet nobody mentioned and an approver who signs off from her phone in a Slack thread, and neither shows up in a requirements document. Both decide whether an agent works.

So the first on-site days are mostly listening. I ask the person doing the work to walk me through yesterday instead of the process document, and I note every time they open something the official flow doesn't mention.

OpenAI and Palantir build that time into the job: OpenAI's FDE postings mention up to 50% travel. Palantir expects its FDEs to spend around 25% of their time on site, according to The Pragmatic Engineer.

Remote is fine for the building. Discovery needs a chair next to the ops lead, and so does the first time they try the new workflow and frown at it.

In five FDE listings I checked on Built In NYC, advertised pay in October 2026 ran from $135,000 at the bottom of Draftwise's range to $369,000 at the top of Legora's.

These are advertised ranges, not offers. Indeed's US average base for the role is about $179,000 and Lightcast's median advertised pay is about $188,000, so these listings run from below the national middle to well above it.

A New York hire also carries a recruiter fee of around 25% of first-year base, Paraform's stated fee, and an 8-12 week search. The full cost breakdown covers what contractors and studios charge.

Companies where AI has to run inside systems they don't fully control, and where nobody on staff is free to sit with the people doing the work.

AI startups selling to enterprises. The product works in your demo, then it meets a customer's data in a system older than your company, and someone has to sit with that customer's team until it works there too. At seed or Series A that job falls to a founder who should be selling, and Paraform says 59% of the companies hiring FDEs on its marketplace are at that stage.

Fintech and legal tech. A wrong answer costs more when it moves money or changes a contract. These teams need a human approval step wherever money or risk is involved and an audit trail for every action the system takes, and both are cheaper to build in at the start than to bolt on after the security review.

Ops-heavy companies with a stalled pilot. The pilot worked in a sandbox and stalled at the system of record. You'll hear that 95% of AI pilots fail, a line from MIT NANDA's 2025 report, which leans partly on 52 interviews and 153 surveys collected at conferences and which Wharton's Kevin Werbach publicly challenged. Less quoted is the same report's finding that projects built with outside partners succeeded about twice as often as internal builds, about 67% against 33%, self-reported.

Where it doesn't fit. A standalone app with no system of record behind it needs a builder more than someone in the room, and my AI consultant page covers that kind of project.

I'm at your office on the days we agree and remote for the rest of the week, with the days and the first 30-day goal set before I start.

A good first 30-day goal names one workflow and one result, like getting invoices from your biggest suppliers coded and routed in shadow mode. "Explore AI use cases" is a goal I'll push back on.

On-site days go to the people who'll use the system. They see early versions inside their own workflow, long before anything is finished, so their feedback arrives while it's still cheap to act on.

Remote days go to building inside your systems and writing the evals that tell us whether it works. Before go-live, the new workflow runs in shadow mode next to the old one so the team can compare results. The engagement ends with a written handover and your team running the system while I'm still around for questions.

I've been building software for 11+ years and have built 10+ AI products end to end. At Chatterbox, a corporate language-learning company, I took over as lead engineer and rebuilt the platform through a seed round and the B2C to B2B pivot. Earlier, at NTT DATA, I rebuilt Sky Italia's "Fai da Te" app and its store rating went from 3.0 to 4.5. I also speak Italian and Portuguese.

If you're weighing an FDE against a consultant, the FDE vs consultant comparison lays out who does what. For technical leadership more than hands in your systems, I also work as a fractional CTO in New York.

My interim engagement is $16,000-$20,000 a month for 2-3 days a week, and scoped pieces of work are $175-$250 an hour.

I don't take full-time W-2 roles. If the job is permanent, hire, and use the Built In NYC ranges above as your budget. When one workflow needs to work this quarter, an interim FDE skips the recruiter fee and the 8-12 week search, and my guide to hiring an FDE walks through both routes.