
Forward Deployed AI Engineer, Backend
Posted 7 days ago
The role in short
Embeds with law firms to model their precedent into Draftwise's legal ontology and build the backend services and LLM pipelines behind it.
About Draftwise
Draftwise is a contract drafting, review, and negotiation AI powered by your legal knowledge. We turn a firm's guidance, precedent, and language into a legal ontology, ground our models in it, and run complex drafting, review, and search workflows directly within Microsoft Word and in the browser.
We serve top law firms and legal departments globally, including over half the Vault 10, dozens of Am Law 100 firms, and Fortune 500 organizations. We are headquartered in NYC with offices in London and West Palm Beach, are Y Combinator alumni, and raised a $20M Series A from Index Ventures.
The role
A Forward Deployed Engineer at Draftwise embeds with a firm, learns how their lawyers actually draft, and extends the product until it works the way they work. You will model their precedent and playbooks into our ontology and write the backend behind it, and when the platform cannot do what a firm needs, you build it.
This is the early career version of that job, weighted toward backend engineering and the AI systems underneath the product. You will learn forward deployed engineering and ontology modeling from engineers who built those practices at Palantir, with real customer accounts and real code from your first month.
The scope is broad. You get ownership and impact across the technology, the product, and our culture, working alongside a first rate engineering team to ship new features, scale the platform, and make it more performant.
What you will work on
Ontology modeling for legal knowledge. Every firm has its own precedent, playbooks, and preferred language, and the ontology is how we turn that into structure the product can reason over. This is the most transferable skill in the role.
Backend services behind drafting, review, and search. We are not religious about languages. There is what the codebase was historically built in and there is whatever gets the job done now, and the complexity that lasts sits underneath that: Postgres, OpenSearch, a graph database, and a wide set of AWS services. You will own features end to end and learn how those pieces fit together, which is rarer to pick up early than any one language.
LLM systems in production. You will write and test prompts, build retrieval and evaluation scaffolding, and measure quality on real customer data, so the work looks more like building test infrastructure than tuning prompts by feel.
Scale and performance. We run large documents and large precedent sets under interactive latency, so profiling, query tuning, and cost work show up as regular engineering here.
Deployments into elite law firms, with a senior FDE beside you. You will own pieces of a firm's rollout early, and you will be the one working out why their playbook does not map cleanly onto the product yet.
Enough frontend to finish the job. Our surface is a Word add in and a web app in TypeScript and React, and we expect you to ship a UI change without waiting on someone else.
What we are looking for
High ownership and low ego. Most of our work arrives as a customer problem instead of a spec, so we look for people who take it off someone else's plate, ask for feedback early, and do not care who gets credit for the fix.
A focus on user impact. We want someone with a passion for shipping high quality products that delight users, which here means a lawyer finishing a markup faster and noticing why.
Curiosity and intelligence over credentials. We hire on how you reason about data models, failure modes, and tradeoffs, and you will be in unfamiliar territory often, legal concepts, an ontology, a service someone else wrote, so the learning has to be part of the appeal.
Backend as your center of gravity. You should already be comfortable with SQL, API design, and debugging a service you did not write, and you should want to get much better at all three.
Comfort in the room with the customer. You will be on calls with partners and knowledge management leads, prepared alongside a senior FDE, and the skill is asking a sharp question and saying plainly what you still need to go find out.
The instinct to turn a vague problem into a model. Lawyers describe problems in terms of their work, and the job is to hear "our fallbacks never come out right" and get to a schema, a retrieval change, or a bug.
Interest in LLMs and their limits. We use Claude Code and similar tools every day, and what matters is the judgment to see where they help and where they quietly cost you time.
You do not need legal industry experience, production LLM experience, or prior customer facing work, and we care that you have built real backend software rather than which language you built it in. Bring the foundation and the curiosity, and we will teach you the rest.
Nice to have
Project or internship work on retrieval systems, evaluation harnesses, or agent frameworks.
Any experience in front of customers, including consulting, support, or solutions work.
Exposure to legal, financial, or other document heavy domains.
Anything you shipped to real users, personal projects included.
Why this role
The real FDE experience. You will learn deployment and ontology work directly from engineers who did it at Palantir, alongside legal professionals from top global firms, and you will do the work rather than watch it.
Proven product market fit. Lawyers at Vault 10, Magic Circle, and Am Law 100 firms use Draftwise for their most important work, so you are building on demand that already exists.
Immediate feedback. Your work directly changes how legal professionals serve their clients, and you will often hear about the impact from the customer within days.
Ownership that compounds. Forward deployed work puts you in front of the customer, the commercial reality, and the code at the same time, which is rare this early in a career. It is the fastest way to become an engineer who can carry a problem end to end, and that travels with you wherever you go next, including a company of your own.
Early stage, real equity. You are joining at Series A, when the work you do still bends the company's trajectory, and every engineer holds a meaningful equity package, so you own a piece of what you build.
Process
Intro call with the hiring manager.
Technical conversation on a system you have built.
screen share coding exercises of about an hour on backend, data modeling, debugging
Take-home project of about 3 hours on a self contained problem, using our stack and the AI tools you would use on the job. We share the rate and the evaluation criteria in advance.
Final conversation with a founder.
Prep for this interview
5 questions likely to come up at Draftwise, each matched to a line in this posting.
- 1Integrations and APIs
They wrote: “Backend as your center of gravity. You should already be comfortable with SQL, API design, and debugging a service you did not write, and you should want to get much better at all three.”
A customer's API starts returning errors in production at 2am. What do you do?
How to answer
Why they ask: FDEs often own production for their customers. This tests incident handling and communication.
- Check your side first: logs, recent deploys, retries piling up
- Stop the damage (pause jobs, queue requests, turn off the feature)
- Tell the customer what you see and what you've done, in plain words
- Afterwards: add a check that would have caught it, and write a short report
Tip: Split the answer into the first hour and the week after.
- 2Data and SQL
They wrote: “screen share coding exercises of about an hour on backend, data modeling, debugging”
The customer's data is spread across five systems that disagree. How do you get one clean view?
How to answer
Why they ask: Real deployments rarely have one source of truth. They want a practical approach, not a perfect one.
- List each system and what it's the best source for
- Pick matching keys and decide how to handle records that don't match
- Load into one place with clear rules, and keep the raw copies
- Show the customer the mismatches; they often fix the source
Tip: Name the system the business trusts most for each field, and get the customer to agree.
- 3LLMs, RAG and agents
They wrote: “Interest in LLMs and their limits. We use Claude Code and similar tools every day, and what matters is the judgment to see where they help and where they quietly cost you time.”
A customer wants a chatbot over their internal documents. How do you build it?
How to answer
Why they ask: RAG is the most common first project. They want the whole path, from messy files to answers people trust.
- Collect and clean the documents; handle permissions per user
- Split them into chunks, embed, and store with source links
- Retrieve, then answer with citations; say “I don't know” when retrieval is weak
- Build a test set of real questions with expected answers before launch
Tip: Spend more time on the data and on testing than on the model.
- 4Coding
They wrote: “Enough frontend to finish the job. Our surface is a Word add in and a web app in TypeScript and React, and we expect you to ship a UI change without waiting on someone else.”
Here's a CSV export from a customer. Write a script that cleans it and loads it into our API.
How to answer
Why they ask: Very close to real FDE work. They watch how you handle bad rows and failures.
- Read with a proper CSV parser; check the header first
- Validate each row and collect errors, not crash on the first one
- Send in batches with retries and a clear report at the end
- Write one or two quick tests for the parsing
Tip: Talk through edge cases out loud: encodings, missing fields, duplicates, rate limits.
- 5Ownership and working with customers
They wrote: “The scope is broad. You get ownership and impact across the technology, the product, and our culture, working alongside a first rate engineering team to ship new features, scale the platform, and make it more performant.”
Tell me about something you shipped that broke in production.
How to answer
Why they ask: Everyone has broken production. They want to hear you own it and change something afterwards.
- What broke and who it affected
- How you found out and what you did in the first hour
- The root cause, said plainly
- What you changed so it can't happen the same way again
Tip: Pick a real one with a cost. Spend most of the answer on what you changed.
Not the one? Get the next one by email
New roles land here every day. Pick what you want to hear about.
Want the job, not just the posting? Learn how FDEs work.
FDE Info is our free weekly newsletter for forward deployed engineers: tools, field notes from real deployments, and how FDE teams hire and pay.
Free. One email a week. Unsubscribe in one click.
