Senior Applied AI Engineer (Agentic Systems)

About the Role This is a founding-level, hands-on engineering role at a seed-stage AI startup in the enterprise compliance automation space. The company builds AI agents that review marketing and packaging content against brand guidelines and market regulations — turning hours of manual, high-stakes compliance workflows into minutes of automated precision for global, regulated enterprises. As a Senior Applied AI Engineer focused on Agentic Systems, you will own the core technical moat: the deterministic orchestration layer and safety infrastructure that makes enterprise-grade AI compliance trustworthy at scale. You're not building generic automation — you're building the category-defining multi-agent infrastructure for a market where correctness is non-negotiable. This is an in-office role in San Francisco with WFH flexibility, working closely with a small, senior team where your work directly shapes product architecture and customer outcomes. What You'll Do Agentic Reasoning & Orchestration: Design and evolve multi-agent LLM systems that decompose complex review tasks into reliable, auditable steps. Define agent responsibilities, hand-offs, and termination conditions to minimize reasoning drift and maximize consistency. Context, Retrieval & Memory Systems: Architect retrieval pipelines using RAG, structured memory, and graph-based retrieval approaches to surface the right context (brand guidelines, regulations, historical decisions) at the right time. Balance recall, precision, and latency across large knowledge bases. Stateful, Asynchronous Workflows: Own long-running, fault-tolerant workflows using Temporal or similar frameworks — ensuring retries, versioning, and determinism across non-deterministic model calls. Treat agent orchestration as a distributed systems problem. Evaluation, Safety & Reliability: Build evaluation frameworks using statistical metrics, gold labels, and automated regression testing to prove system reliability — especially in high-risk legal and compliance scenarios. Asset Understanding Pipeline: Collaborate on image and document preprocessing (OCR, layout analysis, vision-language models) to ensure downstream agents receive structured, machine-readable context from marketing and packaging assets. Cross-functional Collaboration: Partner closely with product, engineering, design, and compliance teams to translate real-world enterprise requirements into robust technical solutions. What We're Looking For 7+ years of professional software engineering experience, with a clear track record of career progression. Demonstrated experience building and deploying multi-agent LLM systems or agentic AI architectures in production. Strong command of RAG pipelines, vector stores, and retrieval system design (graph-based retrieval a plus). Hands-on experience with stateful workflow orchestration tools such as Temporal, Prefect, or similar. Solid understanding of distributed systems concepts: state management, fault tolerance, observability, and versioning. Experience designing LLM evaluation frameworks with quantitative rigor (not just vibes-based review). Familiarity with vision-language models, OCR, or document understanding pipelines is a strong plus. Startup experience strongly preferred; candidates with deep background in CPG, consumer health, or other regulated industries will also be considered. Strong cross-functional communication skills — ability to work across product, engineering, and compliance stakeholders. Minimum average tenure of 2 years across prior roles. Note: Visa sponsorship is not available for this role. Candidates must be authorized to work in the United States. Compensation & Benefits Salary: $185,000 – $210,000 USD annually, depending on experience. Early-stage equity in a company at a significant inflection point. Opportunity to be a founding technical voice shaping architecture, culture, and hiring. Location This role is based in San Francisco, CA. The team works primarily in-office with WFH flexibility. Remote candidates will not be considered.