A senior partner picks up a marker and sketches how she decides whether a client needs a particular regulatory filing. Three questions go down the left side of the whiteboard. Two exceptions hang off the second question like footnotes. At the bottom sits a box labelled "depends on how the board feels about risk". Turning that sketch into logic a system can run, and knowing which box must stay with a human, is what this role is about. Filed under legal tech among remote legal knowledge engineer jobs, this one is full-time, fully remote, available from any country and paid $125,000 per year in USD.
Put plainly, the legal knowledge engineer job description here is to translate legal processes into automated workflows and document assembly in Docassemble, build and test AI legal tools with Python, write prompt and logic specifications, and work with lawyers and engineers on product fixes. The sections below follow a typical project.
Most legal expertise lives in people's heads and old emails. Your first task is to get it out, through structured interviews with several practising lawyers, because two partners in one group rarely apply a rule identically.
Good capture sessions feel closer to cross-examination than a chat. What triggers the question? Which facts change the answer, and which are irrelevant though clients always mention them? What happens when information is missing? When an answer begins with "usually", ask about the unusual case. Collect the precedents, checklists and clause banks they rely on.
Next you model it. Linear processes fit a flow diagram; branching decisions with nested conditions need a decision table or tree, which makes gaps visible. The classic finding is the missing branch: the client is a subsidiary, or the deadline has passed, and nobody said what happens.
This stage also settles scope. Rule-based steps can be automated. Steps involving discretion or professional judgement should route to a lawyer with the facts already gathered. Getting supervising lawyers to sign off that boundary protects client and firm.
Docassemble, the main build tool here, is an open-source platform for guided interviews that collect facts and assemble documents from templates. Interviews are written in YAML with embedded Python; templates are usually Word files with variables and conditional text. You will:
A good interview produces output the reviewing lawyer would not redraft, and never leaves the user guessing what a question means.
Python carries what YAML cannot. Eligibility calculations, date logic and anything with more than a handful of conditions belong in plain modules that can be tested alone. You will also write integrations: pulling matter data from a practice management system, filing documents to a document management system, or calling an AI model's API.
A wrong branch produces a wrong legal document, so tests matter. Expect unit tests for each rule and scenario tests that run an interview end to end with known facts against a lawyer-approved output. Git is assumed.
Legal AI tools such as Harvey can draft, summarise and answer questions over documents, but need direction. You write prompt and logic specifications stating what the tool should do, which sources it may rely on, what it must refuse and what a correct answer looks like.
Then you evaluate. Build test questions with lawyer-agreed answers, run them, and grade outputs for accuracy, sourcing and appropriate hedging. Guardrails follow: limits on what the tool answers, human review before anything reaches a client, clear labelling of AI drafts. Confidentiality and professional responsibility rules apply and vary by jurisdiction, so the supervising lawyers stay responsible for the advice.
A workflow correct last quarter can be wrong today. Statutes are amended, regulators change forms, courts reinterpret tests and firms revise policy.
You keep a register of the rules each tool depends on and who owns them, set review dates, and rerun the suite whenever a source changes. When a lawyer flags a fix, trace it to the rule, change the logic, update the tests and record why, so another engineer or an auditor can follow the reasoning.
Remote collaboration is mostly written. You work with product managers on priorities, software engineers on deployment and lawyers on sign-off, often across time zones. Specifications, decision tables and recorded walkthroughs stand in for a lot of meetings; a week mixes a few video calls with long stretches of build time, tracked in tickets and a shared repository.
Of all the legal knowledge engineer skills, the one that matters most in work from home teams is translation both ways: telling engineers why a legal exception cannot be simplified, and telling lawyers why a requested feature would make the tool harder to test.
This role asks for a law degree or a legal practice background combined with strong technical skills, typically 5+ years across law and legal tech. Proficiency in Python, Docassemble and legal AI tools like Harvey is expected, along with solid process design ability and excellent collaboration with product teams.
Legal knowledge engineer interview questions usually probe both sides: break a familiar process into conditions, read some Python, describe the tests you would write for a branching rule, critique an AI output. A strong legal knowledge engineer resume shows shipped work, such as a Docassemble interview, a repository or an evaluation set, with the problem each solved.
Many of these employers work to US practice; any bar admission, licence or location condition will sit in the posting itself and differs from one to the next.
For this post the figure is $125,000 a year in US dollars, full-time, from wherever you live. Any legal knowledge engineer salary elsewhere will vary with seniority, employer and region, so treat this number as one data point rather than a market rate.
Listings for legal knowledge engineer jobs remote worldwide tend to come from legal tech companies building automation and AI products, or from law firm innovation teams, in-house legal operations groups, alternative legal service providers and access-to-justice organisations building public guided interviews. Each sets its own hours and terms, so read every offer closely.
Most people here started as lawyers who kept automating their own work. Lawyers curious about how to become a legal knowledge engineer tend to follow one order: learn Python properly, build small Docassemble interviews for a process you know, then take automation projects at your organisation. Legal tech clinics, open-source projects and pro bono automation build a portfolio.
No single legal knowledge engineer certification is widely recognised. A legal knowledge engineer course or Python class helps, but working examples usually count for more. Without law or software experience this role is out of reach; junior legal operations or automation analyst roles are a realistic first step.
Employers use these titles loosely. A legal knowledge engineer usually focuses on capturing expertise and encoding it as rules, interviews and AI specifications. A legal engineer leans further into software, building and deploying systems. A legal technologist is broadest, covering tool selection, implementation and training with less coding. The duties list tells you more than the label.
Legal AI has made specification and testing more visible. Firms adopting AI need people who can define correct, design evaluations and decide where a human stays in the loop. That need looks durable, though no role is guaranteed and the tools will keep changing.