Original job description
Forward Deployed AI Engineer
Intermediate · Full-time
Location: Kitchener-Waterloo, ON (in-office up to 3 days/week)
Experience: 3+ years
Eligibility: Authorized to work in Canada
About Arkanis
We're a production AI systems shop. Growth-stage companies bring us in when they have a real engineering problem and not enough senior bandwidth to solve it well: data infrastructure that's buckling, an LLM feature that breaks the moment real users touch it, a matching engine that needs to actually work. We embed with their teams, ship the system, and harden it for the long haul. Our work spans LLM agents, data pipelines, and applied ML across a rotating set of clients.
Who You'll Work With
Neither founder came up the standard way. Both started in the hard sciences and taught themselves computer science and machine learning from there. They worked with the first generation of transformer models, shipped some of the earliest commercial LLM applications running inside Fortune 500 companies, led engineering teams at unicorn AI startups, and put research into production at world-class labs including Microsoft Research and DARPA. None of it came with an Ivy League stamp or a FAANG badge. We earned it by shipping, and that's the only resume line we actually read, and the only thing we test for. No leetcode theater in our interviews, just real systems and whether you can build and reason about them.
The Role
Forward deployed means you're embedded. You'll sit inside a client's team as the engineer who takes an ambiguous business problem and turns it into a working, maintainable production system. You'll move between engagements as projects ship, so you'll see more codebases, stacks, and founding teams in a year than most engineers see in five. You're also the technical face of Arkanis in those rooms, which means the quality of your work and your judgment both matter.
This role is based in Kitchener-Waterloo. You'll work from our Kitchener office up to three days a week, with the rest remote, plus time on client sites during active engagements. We're around to pair on hard problems and back you up however you need it. This is a role we expect to grow someone into: with strong fundamentals and real curiosity about applied AI, we'll teach you the production ML and agent engineering on the job, with senior ICs to learn from.
What You'll Work On
No one person does all of this. Think of it as the range our engagements pull from, the kind of work you'll get to lean into and grow across over time:
Setting up the testing backbone that turns 'move fast and break things' into 'move fast and keep them working.'
Building the testing and CI that let us ship AI features fast and sleep at night.
Designing thoughtful evals and benchmarks, model guardrails, experiment design, observability, and A/B tests, so we actually know whether a feature is working.
Prompt tuning and shaping model behavior until the results are reliable.
Improving stability and performance across the stack for our clients.
Bringing product sense and UI/UX instincts to how a feature feels to use and whether it solves the real problem in front of the user.
Turning a fuzzy business problem into a scoped plan you can ship against, and being honest with founders about what's possible, what isn't, and what's worth doing now versus later.
Documenting and handing off cleanly, so the client's team can own the system after you rotate to the next engagement.
Who You Are
We're looking for builders with staying power. The people who thrive here have gone deep on something hard by their own drive, stuck with it long past the point most people quit, and come out with real command of their craft. Where you learned it and how long it took matters far less to us than the fact that you did, and that you'd keep building whether or not anyone was paying you.
3+ years shipping software. You've built things real people used, debugged them in production, and lived with the consequences of your own design decisions.
Solid engineering fundamentals. You write code others can maintain, you're comfortable across the stack, you reach for version control, tests, and SQL without being told, and you can reason about a system you've never seen before.
A working understanding of what LLMs can and can't do. You know when a problem wants a model and when it wants fifteen lines of deterministic code. You've felt the failure modes firsthand and you're hungry to go deeper on evals, guardrails and design.
Opinions about your tools. You have a coding setup you've thought about and can defend, and you're the kind of person who tries the new thing and forms a real take on it.
Strong communication. You can sit across from a non-technical founder and make the case for security and stability work in terms they care about, manage expectations honestly, and hold your own in a room with VCs and senior engineers.
Comfort with ambiguity and client-facing work. Engagements rarely arrive as clean specs. You're energized by walking into a half-defined problem and a team you've just met.
Nice to Have
Working experience with our common stack: TypeScript and/or Python, Postgres, cloud platforms (AWS/GCP), vector databases.
Exposure to regulated or compliance-heavy environments is a real plus: GDPR, PIPEDA, SOC 2, PCI-DSS, or hands-on PII handling. Several of our engagements touch sensitive data, and knowing how to build for it from the start is valuable.
Any prior exposure to applied ML, computer vision, or LLM observability tooling.
Startup or agency experience where you wore several hats at once.
Why This Role
You'll get embedded at well-funded growth-stage companies and earn real facetime with the founders, investors, and senior engineers running them. You'll be deliberately trained up in production ML and AI engineering rather than left to figure it out alone. And the variety is the point: you'll work across very different companies and problems, building range and depth far faster than any single team could give you.
