Senior Backend Engineer, Architecture Engineering: Nonlinear Productivity

- Identify sources of friction across GitLab's software development lifecycle and scope agentic solutions to address them, turning vague pain points into concrete, buildable proposals.

- Design and build reliable AI-powered systems that follow step-by-step workflows, use tools and safety checks, and correct errors before taking engineering action - the kind of output you can actually trust with real engineering decisions.

- Build and maintain evaluation tools that judge agent output on correctness, constraint compliance, and cost, not on whether it merely "seems to work."

- Work across GitLab's codebase as each problem requires, going wherever the friction actually is rather than staying inside one service or product area.

- Apply distributed systems judgment to identify generated code that may fail under concurrency, at scale, or across self-managed, dedicated, and multi-tenant deployments, catching failures before they reach customers.

- Collaborate with the India-based group, sharing roadmaps, findings, and reusable agent tooling

- Take ownership of a greenfield problem space from day one, helping shape a proven internal fix into a capability GitLab could offer customers externally, with your scope and impact free to grow as the team scales.

- Hands-on experience building agentic or large language model-based systems - multi-step orchestration, tool use, guardrails, and recovery patterns - and making them reliable in production, not treated as one-off prompts or demonstrations.

- A track record of working autonomously in unfamiliar codebases, getting oriented quickly, and driving solutions through completion.

- Strong distributed systems knowledge, including coordination, consistency, idempotency, rate limiting, failure modes, and degradation under load.

- Proficiency in Go, Rust, or Python, in that order of team priority, and the ability to read and modify code in the other languages.

- Helpful experience includes shipping autonomous agents that complete real tasks from start to finish; improving build systems, release processes, review workflows, or other parts of the software development lifecycle; and working with globally distributed teams, large language model workload costs, or production constraints across on-premises, air-gapped, single-tenant, and software-as-a-service deployments.

Nonlinear Productivity - shortened internally to "NLP," with no relation to natural language processing - is one of GitLab's newest teams: a strategic incubation group that reports into AI Platform leadership under the direct sponsorship of the CTO. It's split into a US group (this role) and an India-based group working the same charter; the two sync on roadmap and tooling a few times a week but otherwise run day to day on their own.

It operates like a startup - no dedicated product manager, no pre-set backlog - and solutions that prove out internally are the team's path to a monetized, customer-facing GitLab offering.

It's a good fit for engineers who'd rather go find the next problem than be handed one, and who want a hand in defining a brand-new part of GitLab from its first commit.