AI Native Builder/Architect

Atlan is building the context layer for enterprise AI. Enterprises are pouring money into AI and most of it dies in production because the AI does not understand the business context around the data. That is the problem Atlan solves.

Gartner has named context the defining issue in enterprise AI, and calls context graphs the essential infrastructure for AI agents, naming Atlan one of three vendors already building it. We are also the only vendor named a Leader across all four major Gartner and Forrester evaluations for data catalogs, data governance and metadata management, the foundation the context layer runs on.

Come build the infrastructure that AI runs on.

Gartner calls context the defining issue in enterprise AI, and context graphs the essential infrastructure for AI agents. It names Atlan as one of three vendors already building one. We're also the only vendor named a Leader in all four major Gartner and Forrester evaluations for data catalogs, data governance and metadata management, the layer our context graph is built on.

Come build the infrastructure AI runs on.

You'll join the team building Atlan's context layer, the platforms that make it possible and the vertical agents that run on it. The platform is what any agent needs before an enterprise will trust it: a registry of which agents and skills exist, traces and evals on every run, and governance over cost and access.

Vertical agents each do one specific job, for customers and for Atlan's own teams. Inside Atlan they already run real work. There's an agent our CSMs trust, three GTM agents that replace 3,000 calls made by hand, an agent that watches every page on atlan.com, and one that monitors the other agents. We write up what we learn in the open at Becoming Frontier.

The team works in small "two-slice" units. You own one initiative, usually with one or two people alongside you, and you decide how it gets built.

You'll take problems nobody has fully defined yet and ship working systems, fast. Depending on where you're matched, you may:

• Own a 0→1 initiative from problem to production. Frame it, write the spec when there isn't one, pick the architecture and build it.

• Build agents with the people who own the outcome, inside Atlan or at a customer. Scope the workflow with them, demo early and put the agent where the work already happens, like an approval in Slack or a redline in a legal team's Word doc.

• Earn autonomy with evidence. Start with shadow runs, study the misses and widen what the agent can do as results support it.

• Design for failure. Decide what happens when a tool is down, context is thin or the agent isn't sure, and how a person takes over.

• Make the engineering calls on what stays deterministic code and what needs a model, build vs. reuse, model choice, cost and reliability. Explain the trade-offs clearly.

• Make agent behavior observable. Tie traces and evals to the revision that ran. Track consistency, cost and latency alongside success rate, and treat drift as a regression.

• Build on and extend the context layer, agent platform and registry: versioned skills, layered instructions, explicit permissions for tools and secrets, and model changes handled as migrations with evals rerun before promotion.

• Build AI-native yourself. Run coding agents on your own work, keep their skills and instructions up to date, and review what they write for architectural drift.

• Share what you learn. Turn what works into skills, reference architectures and quality gates that others adopt, and demo what you've built.

• You've built software systems hands-on and shipped 0→1 products or platforms into production, often without a spec.

• You've shipped agents or LLM systems that real people use. When one went wrong, you traced it back to the context, tool call or decision that caused it.

• You use evals to find and fix real problems, check what a score actually counts, and judge agents by user impact: hours saved, error rate or cycle time.

• You design for failure first: what the agent may touch, what happens when it's wrong, and where a person approves before it acts.

• You build with coding agents every day and hold what they write to the same bar as your own code.

• You're a critical optimist. You believe what agents can do is real, and test what still fails before trusting them with more.

• You reason from first principles. When a familiar design doesn't fit, or a better model makes old scaffolding unnecessary, you notice and rebuild.

• You have strong systems judgment. You can design for production reliability, latency, cost and security, build it yourself, and leave it in a shape others can build on.

• You do your best work owning an initiative with one or two people alongside you, and don't need a large team to feel effective.

• Experience with MCP, A2A, agent frameworks and runtimes, LLM gateways or agent observability, or you've built an agent harness or orchestration layer yourself.

• Experience with knowledge graphs, vector search, retrieval or semantic layers.

• You've built an eval suite from scratch, or run shadow evaluations, LLM grading or staged rollouts.

• You've shipped agents that write to a system of record, with permissions, approvals and an audit trail.

• You've led AI adoption inside an engineering or business team and can show the before and after.

• Founding-engineer, technical-founder or early-stage startup experience, where you built something from nothing.

• A public trail of work, repos, demos, write-ups or talks, that shows how you build.

This role probably isn't for you if you want to manage a large team, want a well-scoped backlog handed to you, would rather build demos than run agents in production, or are still planning to get into AI.

Joining Atlan means being part of a global movement to help data teams do their life’s best work. Here’s what you can expect: