About Tavrn

Tavrn builds AI agents that run the operations work at personal injury law firms. The agents retrieve medical records, call and fax providers, follow up on requests, build chronologies, and draft demands. Most of this used to be done manually by paralegals.

The role

Forward Deployed Engineers work directly with law firms to understand how they operate and deploy our agents into their workflow. Projects often start with an open-ended question from a partner: “How do we cut our intake time in half?”, “How do we run discovery for a 200-plaintiff case without burning out the team?”, “How do we get our records turnaround under two weeks?” You take that question, work it down to something concrete, and ship.

Your day might span a call with a partner at a firm, debugging an agent’s behavior on a live case, building an integration into a case management system, drafting a memo to ops about something the firm flagged, or deciding what’s worth pulling back into the core product. You own projects end to end and you work with minimal supervision.

The work is different from a standard engineering job. The hard part is usually not the code. The hard part is figuring out what the firm actually needs (which is rarely what they say on the first call), deciding what’s worth building vs. running on the existing platform, and shipping something that a partner uses without flinching.

Who does well here

A few traits matter much more than years of experience:

• You test your assumptions. You don’t ship something and hope. You set up a way to know if it’s working.

• You’re good with AI coding tools, and you don’t trust them. You read the diff. You catch when the model is confidently wrong and you correct it.

• You ship a lot. Including nights and weekends when the work calls for it.

• You know when something is actually done. Good enough that a partner at a firm uses it without flinching.

• You’re good with customers. You listen, ask questions, and push back when you have more context than they do.

• You write clearly. Recaps, memos, follow-ups. The work depends on ops, sales, and engineering staying in sync, and that mostly happens in writing.

Reading comprehension and judgment matter more than CS theory. Most of the bad outcomes we’ve seen in this kind of role come from someone building the thing they were asked to build, instead of figuring out what the firm actually needed.

Qualifications

• Graduating 2026 or recent grad.

• Projects, internships, or research where you shipped real software people used.

• Comfortable in Python and TypeScript.

• You’ve used AI coding tools enough to have real opinions about how to use them.

Hiring process

Start to offer is usually around 10 days.

1. 30-minute screen.

2. 2-hour live case study. 15 min scoping a problem with someone roleplaying as a partner at a firm. 90 min building a solution live (the model you’ll be using is small, so we can see how you plan, push back, and verify). 15 min walking through what you built and defending the decisions.

3. 2-day in-person work trial in SF on a real ticket from our backlog.

4. Offer.

How to apply

Email your CV to vitor@tavrn.ai with “Forward Deployed Engineer, New Grad” in the subject.