Senior AI Engineer

GitLab – Senior AI Engineer
About the Role
Here's the honest version of what this job is: you'll spend a lot of time figuring out what not to build.
That's not a knock on the role — it's actually what makes it interesting. Anyone can pick up a framework and wire together an LLM integration. What's harder, and rarer, is the person who sits down with a Sales team, listens carefully to what they're describing, and comes back with "actually, your real problem is a handoff issue between two tools, and we can fix that without AI at all." That kind of thinking is what we're looking for.
When AI is the right answer — and often it will be — you'll own the whole thing. Discovery, design, code, deployment, iteration. You won't hand it off after the fun part. You'll stick around to find out whether it worked.
Your early focus will sit across Sales, Marketing, and Customer Support. These are functions with real workflow complexity, lots of tool surface area, and genuine room for AI to make work faster and less frustrating. You'll embed solutions directly into how those teams operate day to day, not build things that live in a demo environment and never get used.
This is a fully remote role. We work asynchronously across time zones, and we expect you to be comfortable with that — strong written communication, good judgment about when to escalate versus just make a call, and the ability to move work forward without needing someone in a meeting to tell you what to do next.

What the day-to-day actually looks like
You'll start most initiatives by talking to people and mapping how their work actually flows — not how the process doc says it flows. Those are often very different things. You're looking for the real constraint: the thing that, if you fixed it, everything downstream gets easier. Sometimes that's an AI problem. Sometimes it's a missing integration. Sometimes it's a process nobody's questioned in three years.
Once you've diagnosed the real issue, you'll design a solution and build it. We're not precious about timelines here in the sense of "six months to ship." We expect working prototypes fast — days, not quarters. You'll ship something real, put it in front of users, and learn from what happens.
The metrics that matter aren't just adoption numbers. We care about flow — how long does it take for work to move through the system, where does it pile up, what's the throughput. You'll set up the feedback loops that make those things visible and keep improving based on what you see.
You'll work directly in the systems our teams use every day. Salesforce, Marketo, Zendesk, Workato, Glean — you won't necessarily know all of these deeply going in, but you'll get comfortable building across them quickly. Sometimes that means custom code. Sometimes it means a well-configured platform. Sometimes it means a prompt that's better than anything else you could build. You'll develop the judgment to know which is which.
We use our own products. GitLab Duo is part of how we work, and you'll be expected to use it, push on it, and bring honest feedback about what works and what doesn't back to the teams building it.

What we're looking for
You're a real engineer. Not "real" as a gatekeeping thing — real meaning you can take a problem from nothing to production-quality code, independently, and debug the hard stuff when it breaks. Whether that background came from a traditional engineering job, years of building automations, or a decade of side projects doesn't particularly matter. What matters is that the output is good and you can own it.
Your AI knowledge goes past the surface. You've done prompt engineering seriously — not just "I added some instructions to the top of a message," but actually thinking about context window management, output structure, evaluation, systematic iteration when something isn't working. You understand the trade-offs between different model sizes and architectures. You know what RAG is good for and when it's overkill. You've built things with agentic patterns — tool use, multi-agent setups, human-in-the-loop designs — and you know where those approaches tend to break in production.
You've worked across the LLM ecosystem. Anthropic, OpenAI, open-source models. You have opinions about when to use which, and those opinions are based on actual experience, not just what you've read.
You think about safety as part of engineering, not a compliance checkbox. Prompt injection, data leakage, output filtering, access controls — these are things you build in from the start, not bolt on at the end when someone raises a concern.
Systems thinking is natural to you. You look at a messy process and your brain immediately starts asking where the constraint is. You trace problems to their actual causes. You don't skip straight to "what model should I use" before you understand what's actually broken.
You know enough about enterprise business systems to navigate them. You don't need to be a Salesforce admin or a Marketo expert, but you need to understand how these systems fit together and what it takes to build reliably across them. Data models, integration patterns, API quirks — you pick this stuff up fast.
You're good with ambiguity and independent enough to work through it. We're not going to hand you a fully scoped project with a detailed spec. You'll talk to stakeholders, form your own view of what needs to happen, align on outcomes, and then go build it. If that sounds uncomfortable, this probably isn't the right fit.
You think about the user. Adoption matters. An AI solution nobody uses is an expensive failure. You scope MVPs thoughtfully, cut scope when you have to, and keep the person on the other end of what you're building in mind the whole time.

Things that would give you a head start
You've worked inside GitLab or built on top of it. You have a background in consulting, solutions engineering, or a customer-facing technical role where you had to quickly understand unfamiliar business contexts. You've done value stream mapping or worked with Theory of Constraints in some form. You've used low-code tools like n8n, Make, or Workato — not as a replacement for coding, but as part of a pragmatic toolkit. You've worked at a startup or high-growth company where things change fast and you had to keep up. You've helped junior engineers grow.

The team you'd be joining
Enterprise Technology and AI is the team that keeps GitLab running well internally and pushes on how we can operate smarter. We're not a group that chases new tools because they're interesting. We care about understanding the system, finding the real constraint, and building things that actually improve how work flows.
We're fully remote, async-first, and serious about GitLab's values — collaboration, results, efficiency, inclusion, iteration, and transparency. If you've worked in that kind of environment before and thrived, you'll feel at home here.

Compensation
US base salary range: $139,200 to $218,400. Where you land depends on experience, skills, and how you compare to the rest of the team. The range doesn't include equity or benefits, both of which are real parts of the overall package. Sales-eligible roles may also include incentive pay.

Before you talk yourself out of applying
A lot of people read a job description, count the requirements they don't fully meet, and close the tab. Please don't do that here. We've written down what the ideal person looks like, but ideal people are rare and we know it. If this role genuinely interests you and you can do most of it well, apply. Let the recruiters decide. You might be exactly who we're looking for, even if the description doesn't perfectly match your resume.
GitLab is an equal opportunity employer. Every hiring decision is made on merit, full stop.