Staff AI Engineer - Agent Architecture & Behavior

Artisan•San Francisco, CA

29d•$250,000 - $325,000•Onsite

About The Position

Artisan is developing an ambitious new project at the frontier of applied AI, focusing on agentic AI capabilities. This role is for a hands-on technical lead responsible for designing and building the core AI architecture, encompassing agent behavior, multi-agent systems, tool use, context management, and evaluation. The position involves translating promising ideas into robust production software, offering broad technical ownership and the opportunity to shape the fundamental technology. It requires a blend of exploring uncertain problems and performing detailed engineering to ensure solutions are functional and dependable.

Requirements

Proven experience building and shipping substantial agentic systems, with production use or rigorous, reproducible open-source work being more important than specific frameworks or employers.
Deep practical experience with LLM tool use, planning, context engineering, and evaluations.
Experience implementing multi-agent coordination or substantial parallel agent/tool execution, with the ability to explain its failure modes.
Hands-on experience with browser or computer automation in an agentic system, including state observation, effect verification, and recovery from interface or execution path failures.
Excellent software engineering skills in Python, TypeScript, or a comparable language.
Comfort with asynchronous services, state machines, persistence, concurrency, retries, and cancellation.
Understanding of which decisions are best suited for models versus which guarantees must be enforced in code.
Ability to reason carefully about permissions, untrusted inputs, uncertain external outcomes, and human approvals.
Skill in designing meaningful experiments, debugging real system behavior, and explaining improvements and remaining uncertainties.
Capacity for technical ownership of unclear problems, effective collaboration with other engineers, and shipping with urgency and care.

Nice To Haves

Depth in agent memory and retrieval.
Experience with skill acquisition, reinforcement learning, or post-training.
Familiarity with trajectory datasets.
Experience with sandboxed execution.
Knowledge of distributed systems.
Expertise in inference optimization.
Experience with multimodal and voice models.

Responsibilities

Design and implement agent execution loops, planning strategies, tool interfaces, and verification processes.
Translate ambiguous technical requirements into clear system boundaries and functional code.
Build systems for delegation, coordination, context sharing, and result synthesis in multi-agent environments, handling concurrency, conflicting updates, cancellation, and stale results.
Establish when a multi-agent approach is superior to simpler baselines.
Make complex, stateful workflows resilient to interruptions and partial failures through checkpoints, recovery strategies, and appropriate human intervention.
Improve retrieval, context construction, persistent state, and skill representation for agent memory and context management.
Investigate how systems can leverage feedback and prior experience for performance improvement without regressions.
Build realistic evaluations, analyze task trajectories, and use observed failures to drive measurable improvements.
Compare different approaches based on quality, reliability, latency, and cost.
Evaluate AI models and emerging techniques, prototype promising approaches, and make informed build-versus-buy decisions.
Select tools based on problem-solving effectiveness and be willing to replace them when evidence suggests a better alternative.
Set engineering standards, review critical design decisions, and contribute to the implementation of a coherent AI system.
Collaborate with product and infrastructure engineers on production services, integrations, secure execution, and observability.
Own the AI architecture and its effectiveness, sharing implementation responsibilities across the team.