Staff Software Engineer

Start free →

Or apply yourself on the official page →

About the role

The Role  As a Staff Engineer in HRSD, you set the technical direction for model-driven capability across the HR domain, the agentic and conversational experiences that interpret employee, manager, and HR agent intent, reason over profile, case, catalog, policy, and knowledge context, and act on the user's behalf. Your unit of ownership is the architecture, standards, and evaluation infrastructure that let multiple teams ship those experiences safely — not a single feature set.  This is not an ML research role; you do not train foundation models. It is also distinct from traditional full-stack staff work, where systems follow deterministic logic. Two things shape the job: much of the load-bearing logic lives in natural language — instructions, prompts, context, tool descriptions, guardrails — and must be engineered with the same discipline as code; and because behavior is probabilistic, correctness is established through evaluation at scale, not fixed assertions.  What You'll Own  • AI architecture for the domain. How agents are decomposed and composed, where reasoning happens, how context is assembled and bounded, how tools are exposed, and how autonomy is delegated. Own the multi-release calls: model selection and migration, orchestration approach, build-versus-adopt, and the cost/latency/quality tradeoffs behind each.  • The autonomy boundary. Decide as domain policy which HR actions an agent may take, which require a human decision point, and which no agent should attempt. Anything changing pay, employment status, restricted records, or employee-relations matters requires a human in the path by construction — build the mechanisms that make those constraints structural.  • The shared instruction and tool description surface. HRSD ships as product; customers configure, extend, and override this surface on their own instances. Treat it as a versioned contract with upgrade-safe extension points, deprecation paths, and compatibility guarantees.  • Evaluation as infrastructure. Own the golden datasets, multi-turn suites, judge calibration, CI gates, and drift detection that let teams change behavior safely. Extend coverage to HR-specific failure classes: access-boundary violations, cross-scope leakage, jurisdictional and policy-variant correctness.  • Production quality and safety. Observability for containment, hallucination rate, tool-selection error, unsafe action, and injection vectors from user-supplied content entering agent context. Because HR conversation content is itself restricted, design diagnosis that works without exposing what was said.  • AI-assisted engineering standards. Convert ambiguous problems into testable specs; define what accountable agent-assisted delivery looks like for the domain — specification standards, review expectations, verification harnesses — and hold the line on it.  • Hands-on where it matters. The hard integration, the risky migration, the prototype that settles an architectural argument, the incident nobody else can unblock.  • Partner across functions. Collaborate efficiently with product, engineering, design to co-create scalable AI solutions and translate business needs into robust technical designs.  What You Bring  • 7+ years of software engineering, with a record of technical direction adopted beyond your immediate team.  • Direct, hands-on experience authoring agentic instructions and prompts, designing autonomous workflows, and building the evaluation that verifies them.  • Production experience with LLM APIs, retrieval-grounded features, agent orchestration, tool and function calling, and structured output enforcement — including at least one model or orchestration migration carried through production.  • Entitlement-aware data handling: per-user access enforcement in

View the official posting →