AI Engineer

Hilbert•San Francisco, CA

12d•$160,000 - $180,000

About The Position

Hilbert is seeking an AI Engineer to build production-grade AI systems end-to-end, from prototype through pipeline to product. The role owns core components of Hilbert’s AI stack, including agent architecture, evaluation systems, and production tradeoffs among accuracy, latency, and cost. The engineer will ship in a fast-moving startup environment with evolving requirements and clearly communicate technical decisions and tradeoffs to the team. The initial focus is evaluation and testing for agents already serving enterprise customers. The engineer will design reproducible evaluation harnesses, define correctness for multi-step agent trajectories, establish regression gates, and convert production failures into test cases. The scope will expand to retrieval, orchestration, and execution across the AI stack. The role also involves combining retrieval approaches such as RAG and graph-based retrieval; building robust agent workflows that handle edge cases, missing data, and human escalation; and enabling agents to take actions through external-platform integrations and human-in-the-loop checkpoints. The position sets technical standards for agent development and collaborates closely with the founding team and cross-functional partners.

Requirements

At least 4 years of production software engineering experience, including APIs, services, or data infrastructure.
At least 2 years building LLM or agent systems that shipped to production and were adopted by real users.
Experience owning production code with tests, CI/CD, and on-call responsibilities.
Hands-on experience with LangChain, LangGraph, or equivalent agent/orchestration frameworks, including experience working around their limitations.
Clear, confident communication with both non-technical stakeholders and senior engineers.
Strong ownership and initiative; able to identify needed work and deliver it without waiting for tickets.
Comfort with ambiguity and changing requirements in AI product development.
Ability to work at startup speed and distinguish decisions that need careful consideration from those that can be made quickly.

Nice To Haves

Deep experience with retrieval-augmented generation, including hybrid and graph retrieval, chunking and embedding strategy, ranking, and grounding.
Experience with observability for LLM systems, such as Langfuse, OpenTelemetry, or equivalent.
Experience with cost and latency optimization for LLM systems.
Experience with MCP, tool-calling frameworks, structured output, or constrained decoding.
Experience at early-stage startups or in high-growth environments requiring multiple roles.

Responsibilities

Own the agent evaluation layer, including harnesses, metrics, golden datasets, regression gates, and human-in-the-loop review.
Architect and implement agent workflows using LangChain, LangGraph, or equivalent frameworks, including state memory, routing, tool registries, and recovery paths.
Own systems from experimentation through production and operate them after launch, including tracing, monitoring, latency, cost-per-task budgeting, and on-call support for systems built.
Diagnose production failures such as hallucinations, tool misuse, retrieval misses, and silent degradation, and turn them into durable fixes and tests.
Set technical standards for agent work by reviewing designs, defining reusable patterns, and raising the bar across the team.
Collaborate with the founding team and cross-functional partners, communicating tradeoffs, progress, and technical decisions clearly.
Make pragmatic engineering decisions under ambiguity; ship, learn, and iterate.
Develop intelligent retrieval strategies combining RAG, graph-based retrieval, and other approaches.
Build robust workflows that use available context, handle edge cases and missing data, escalate to humans when needed, and avoid silent failures.
Integrate agents with external platforms so they can execute workflows and take actions, with human-in-the-loop checkpoints.