SDE III - Engineering Productivity (AI)

Why This Role Exists

Engineering output is the constraint on how fast we ship against Fortune 500 demand. AI can lift it — but most organizations bolt AI onto unchanged workflows and get a chat window. The leverage is in redesigning the workflow around what AI can now do, putting agents inside CI, review, testing, and incident response, giving them real engineering context, and setting guardrails so engineers move fast without lowering the bar.

The objective is not AI adoption. It is materially higher engineering output, speed, and quality.

What You'll Do:

• Find the bottlenecks: Instrument how engineering time is actually spent — cycle time, review latency, CI failures, test authoring, operational toil, onboarding — and rank problems by recoverable hours.
• Pick the right instrument: Decide per bottleneck whether the answer is AI, automation, conventional software, or process change. Reject AI where it's the wrong tool.
• Apply AI across the lifecycle: AI-assisted development and refactoring, first-pass code review, test generation and suite optimization, automated documentation and release notes, design analysis, automated migrations, AI-assisted debugging, incident investigation, and root-cause analysis.
• Build engineering agents: Agents that diagnose CI failures, investigate production issues, write tests, run dependency upgrades, propose security fixes, analyze PRs, and execute repetitive migrations — human-in-the-loop by default, autonomy earned per workflow.
• Redesign the process: Rework code review, testing, CI, and incident management for a world where AI does the first pass — and define what stays human-owned and where approval is mandatory.
• Build the platform: AI gateway, agent runtime and workflow orchestration, codebase intelligence, engineering context retrieval, and deep Git/CI/observability/ticketing integrations — plus APIs and SDKs so teams build their own AI workflows on it.
• Solve the context problem: Give agents permission-aware access to code, architecture docs, service ownership, deployment state, telemetry, incidents, and standards — and own retrieval quality and freshness.
• Set the guardrails: Quality, security, privacy, access control, approval gates, auditability, and evaluation for AI-generated change.
• Prove the impact and drive adoption: Baseline, experiment, publish results, kill what doesn't move the metric, and mentor teams into the patterns that work.

How We Measure Success:

Not by tools built, experiments run, LLM calls, or chatbot users. By outcomes: reduced cycle time and repetitive work, faster PR review and CI failure resolution, fewer escaped defects, faster incident diagnosis, less operational toil, shorter onboarding, higher deployment frequency and throughput, and real adoption of AI workflows.