Advanced Software Engineer - Ai Surfaces
Posted Sep 30, 2026 · We last checked this listing on Oct 6, 2026
Likely interview questions for this role
Written from this job description, not a generic list. Each one notes what the interviewer is really checking.
About your past
This team ships agents to lawyers every week. Tell me about a time you shipped something fast and later had to go back and fix a shortcut you took.
What they are checking: Whether they can move quickly while staying honest about the debt they create and paying it down.
You'll work directly with the founders. Tell me about a time you disagreed with a product decision from someone senior and what you did about it.
What they are checking: Whether they can push back with evidence on a small team without either caving or digging in.
Tell me about something you built that real users touched, and what you changed after watching them use it.
What they are checking: Whether they learn from actual user behavior, which matters when lawyers are busy and unforgiving of friction.
What would you do
A lawyer says the agent cited a passage that isn't actually in the evidence. How do you track down why that happened and stop it from happening again?
What they are checking: How they debug hallucinated citations when accuracy is the whole product and trust is easy to lose.
A partner at a big firm asks for a feature that only their firm wants. How do you decide whether to build it?
What they are checking: Whether they can weigh one loud customer against what makes the product better for everyone.
You're early in this role and the codebase was built quickly by a few people. How do you get productive in the first few weeks?
What they are checking: Whether they can ramp up on an unfamiliar, fast-moving codebase without needing heavy hand-holding.
About the work
Gavel lives inside Microsoft Word. Walk me through how you'd build a feature where an agent proposes edits to a lawyer's draft without wrecking their formatting or tracked changes.
What they are checking: Whether they understand the constraints of working inside a host app like Word instead of a blank web page.
How would you set up evaluations for an agent that drafts and edits legal documents, so you know a prompt or model change made things better and not worse?
What they are checking: Whether they treat LLM behavior as something to measure with real test sets rather than eyeball.
Explain how you'd design the backend for a long-running agent task, like reviewing a hundred-page contract, so the Word add-in stays responsive and the user can see progress.
What they are checking: Whether they can handle streaming, async jobs and state across a full-stack system, not just the model call.
Lawyers work with privileged, confidential documents. What would you build or check to make sure client data stays where it should when you're sending text to a model?
What they are checking: Whether they think about data handling and confidentiality as an engineering requirement in legal work.
How do you decide when a task should be handled by a deterministic function versus handed to the model?
What they are checking: Whether they know where LLMs add value and where plain code is cheaper, faster and more reliable.
Practice this interview out loud.
Offer builds a real interview for this exact role at Relativity from your resume and this job description, asks the questions one at a time, and tells you what landed. The first one is free.
Practice this out loudThe full job description
As published by Relativity.