AI Engineer

About Us

At Lexi, we’re revolutionising how legal work gets done. We are building AI legal associates that complete real legal work end to end. Not autocomplete, not copilots, but associates that can plan, use tools, reason across documents, and work the way a real lawyer would.

For decades, legal work has been bottlenecked by human hours. AI is more than autocomplete now. It can produce leverage across research, drafting, review, and matter work, and we believe AI legal associates will become the default. Lexi is building the company that defines what that looks like.

To learn more about Lexi, what we’re building, and the kinds of people we’re looking for, visit getlexi.io/careers

The Role

As an AI Engineer, you’ll build the systems that power Lexi’s AI legal associates across retrieval, document understanding, drafting, review, and multi-step legal workflows. This role matters because Lexi is not building a chatbot for lawyers. We are building agents that need to reason across documents, coordinate across workflows, produce high-quality outputs, and perform reliably in production environments where accuracy, latency, and trust actually matter.

What you'll own

• ▸Building LLM-powered agents and multi-agent orchestration systems that can complete complex legal tasks across documents, matters, and user instructions.
• ▸Improving retrieval, document processing, and context management so Lexi’s outputs are more accurate, grounded, and useful.
• ▸Designing backend systems, APIs, and services that are secure, reliable, observable, and built for real customer usage.
• ▸Creating evals, benchmarks, guardrails, and instrumentation that help us measure and improve agent quality quickly.
• ▸Working closely with founders, product, and legal users to turn messy workflows into systems that actually ship.

What we're looking for

• ▸You have strong engineering fundamentals and can write clean, reliable production code in Python.
• ▸You have hands-on experience building with LLMs, agents, multi-agent orchestration systems, retrieval systems, or AI-powered workflows.
• ▸You are comfortable working with Postgres, Redis, queues, Docker, cloud services, and async pipelines.
• ▸You care about evals, benchmarks, latency, observability, and reliability, not just getting a demo to work once.
• ▸You can take an ambiguous product or workflow problem and turn it into a working system without waiting for perfect specs.
• ▸You have a strong product mindset, move fast, and take ownership of what you ship.

Who thrives here

We’re a lean team taking on one of the oldest industries in the world. The people who do well here are high-agency, direct, and comfortable building where there is no playbook.

• ▸High ownership: You own outcomes, not tasks. No one will hand you a perfectly scoped problem.
• ▸Speed with depth: You move fast without lowering the standard. We care about pace and quality.
• ▸Directness: You give and take feedback clearly. Ego costs time.
• ▸Comfort with ambiguity: You can make progress before everything is figured out.
• ▸Raise the bar: You push for the best version we’re capable of, then better next quarter.

How to apply

Apply using the button below. We read every application ourselves and reply personally to anyone we’d like to learn more about. The best applications show evidence: what you’ve built, owned, improved, or figured out from scratch.