AI Systems Architect
About The Position
We are building a layered AI intelligence system — a multi-layer agent architecture with a dynamic router, context engine, execution loop, verification layer, and eval feedback cycle. This system will be designed to handle long-horizon business tasks that cannot be accomplished in a single model inference. As the AI Systems Architect, you will own the design of the entire system. You will decide how intelligence is structured across layers, how context flows between components, when the system routes to a human, and how failures feed back into improvement. You will set the standards the rest of the team builds to.
Requirements
- Deep understanding of how LLMs behave, fail, and can be steered through prompt design and context structure
- Experience building and shipping multi-step agentic systems in production — not prototypes
- Systems thinking: ability to design interfaces between components, reason about failure modes, and make architecture decisions that scale
- Strong opinions about what does and does not work in LLM system design, backed by real experience
- Comfort with ambiguity — this is a new field and there are no established playbooks
Nice To Haves
- Experience with RAG architectures, vector databases, and retrieval systems
- Familiarity with MCP, LangChain, LlamaIndex, or similar agentic frameworks
- Background in distributed systems or backend infrastructure
- Experience designing eval systems and benchmarks for LLM outputs
Responsibilities
- The router: the component that selects which role to invoke next — executor, planner, verifier, clarifier — based on current task state, risk level, and uncertainty
- The context engine: RAG pipelines, structured memory, MCP tool connections, and the prompt library that gives the system company-specific knowledge
- The execution loop: the action-observe-act cycle that lets the system pursue multi-step goals with real-world grounding
- The verification layer: checker model design, confidence thresholds, and human escalation logic for irreversible actions
- The eval loop: the feedback cycle that makes the system improve over time without retraining the base model
Benefits
- Medical, Dental, Vision, Life, Disability
- 401(k) Retirement Plan
- Unlimited Discretionary Time Off
- 10 paid holidays per year
- 80 hours per year
Stand Out From the Crowd
Upload your resume and get instant feedback on how well it matches this job.
What This Job Offers
Job Type
Full-time
Career Level
Mid Level
Education Level
No Education Listed
Number of Employees
501-1,000 employees
