Principal Engineer - Data

For twenty years, "using the data" meant a person opening a dashboard and clicking until the number showed up.

That's ending. The thing asking the question now is often an agent — a salesperson's AI fetching its own answer, a developer's agent interrogating build history from inside the product. The dashboard was the interface built for humans. What comes next has to work for machines and humans at once, and almost no one has architected a data platform for it.

You'd be the one who does it here.

🔧 The problem you'd own

Buildkite's data goes deep — the build and test history behind how teams like Airbnb, OpenAI, Canva, Anthropic and Pinterest ship software. Storing it was never the hard part. The hard part is the question underneath everything now: can an agent find it, trust it, and act on it — safely, and fast enough to matter?

Answering that means rebuilding from the query engine up. What an agent actually needs from a semantic layer. What "self-serve" means when the person asking isn't technical. Where the current architecture has to bend to make it possible — starting with a live call on moving from Athena to ClickHouse, which you'd be the one to settle.

And the surface is wide: the same platform has to answer a GTM teammate typing a plain-English question and a customer's agent querying build data inside the product. Internal and external, off one foundation.

This is a principal-level role that sets the direction on how this gets solved.

🚀 The work itself

You'd own data architecture at Buildkite: the strategy, the technical calls, and the patterns the rest of the data effort builds on. Most senior data practitioner in the company, and the advocate for data across engineering.

You'd choose the stack rather than inherit it — query engines, semantic layer, ingestion and streaming, storage — and stand behind the choices. You'd design for agents and humans both, so data is discoverable, trustworthy, and safe to act on. And you'd stay hands-on where it counts, chasing the bottlenecks and constraints yourself rather than handing them down. Product and leadership come along through your evidence and trade-offs, not your title.

✨ Who this is for

A few things matter more than any particular tool:

• You've led data architecture at scale — and owned the decisions, not just contributed to them. You have strong opinions on how these platforms should be built, and the track record to defend them.
• You know the modern stack end-to-end — semantic / headless-BI layers, query engines and analytical stores, transformation and modelling, ingestion and streaming — well enough to reach for the right tool over the familiar one.
• You've wrestled with how agentic AI consumes data, or you're hungry to. What a semantic layer owes an agent; what self-serve means when the user is a machine.

Worked in dev tools or CI/CD? Comfortable in a monolith product environment (little Ruby to write, but no fear of it)? Know Kafka or Flink? Build with privacy and compliance in mind? All a head start — bonus, not bar.

The one line we won't budge on: you've owned data architecture in production and made the calls yourself — not advised on them from the side.

🧭 Is this you?

You'd be a strong fit if you:

• Want to own the data architecture direction outright — set the patterns rather than inherit them.
• Get energy from an open problem with no template to copy, and are happy making the call when the answer isn't obvious.
• Want to be the definitive technical voice on data and influence through evidence, not a management title.
• Do your best work async and remote, with a lot of autonomy and not much looking over your shoulder.

Probably not the right role if you:

• Want a people-leadership path — this is a pure architecture role, for the immediate future.
• Prefer a mature data org with settled patterns and plenty of scaffolding to lean on.
• Want the direction handed to you. Here you set it, which means ambiguity and decisions that are yours to own.

None of that is a filter for its own sake — we'd rather you weigh it up now than find the mismatch three months in.

💚 Why Buildkite

• Frontier work. Agentic data isn't a roadmap slide here — it's the problem in front of you, at the edge of where data platforms are heading.
• Real scale, real stakes. The data you'd shape sits in the critical path of some of the strongest engineering teams on the planet, shipping to over a billion daily users.
• Ownership, not tickets. Flat, high-autonomy. Being the most senior data person here is influence you don't get where the function's buried three layers deep. Plainly: less scaffolding, and the direction is yours — which cuts both ways.
• Remote, properly. Since 2013 — async, built for deep focus, with genuine team overlap.
• Small enough that it counts. ~150 people, no hiding. What you build is visible. Bring the weird while you're at it — we hire the whole human, not the role.

What happens next?

Every application gets a response. If this is the problem you've been wanting to get your hands on, apply, or reach out with questions first.