AI Domain Architect

We are hiring the architect who teaches AI what our business actually means. You will be a member of the AI Domain team, the group that owns the enterprise standards, reference architectures, approved model catalog, and governance safeguards for AI across the company. Within that team, you own the domain intelligence layer: the ontology, the agent workflow models, the fine-tuning strategy, and the intelligence pattern library that product teams build on. This is a hands-on senior individual contributor role. Patterns here are proven through working proof-of-concepts before they become standards, so you will build as much as you write.

Responsibilities Platform ontology design

• Design and own the canonical ontology as a semantic layer spanning all PrismHR platforms, anchored in the central PEO and HCM platforms - entities, relationships, business rules, and terminology covering co-employment, payroll, benefits administration, workers’ compensation, onboarding, and compliance - built in partnership with the product teams who own each platform’s data model
• Ground the team’s reference architectures in that ontology so retrieval, reasoning, and generated output stay consistent with how the business and its regulations actually work
• Establish governance for ontology evolution as products, regulations, and customer needs change, and maintain it as a living artifact with a defined review cadence AI agent workflow modeling
• Model business processes across the PrismHR platforms - PEO and HCM foremost - as agentic workflows: task decomposition, tool boundaries, decision points, escalation paths, and human-in-the-loop checkpoints mapped to the team’s risk-level requirements
• Define where agents may act autonomously and where they must defer, especially in payroll, money movement, and compliance-sensitive operations
• Ensure workflow models inherit the team’s embedded safeguards - validated identity claims for tenant context, audit logging, content safety - rather than reinventing them per use case

AI fine-tuning strategy

• Own the strategy for when we fine-tune, when we use retrieval, and when prompting is enough, based on cost, accuracy, and maintenance trade-offs
• Define dataset curation standards for domain-specific training data, including labeling, provenance, and the data-handling constraints (retention, residency) our compliance posture requires
• Build domain-specific evaluation harnesses that feed the team’s model selection and evaluation process, measuring accuracy in PEO, HCM, and the surrounding platform domains rather than generic benchmarks AI intelligence pattern library
• Build and maintain the reusable library of AI intelligence patterns - retrieval strategies, reasoning templates, agent scaffolds, validation guards - so product teams start from something proven
• Validate each pattern through a working proof-of-concept before it is published as a standard, and document when to use it, known failure modes, and evaluation criteria
• Run the feedback obligation: turn implementation lessons, pattern gaps, and production findings from product-team engagements into updated patterns and standards Partnership
• Engage with product teams from design through go-live - reviewing designs, pairing on integration, resolving pattern gaps - while the product team owns delivery and operations
• Advise on use-case feasibility, risk, and fit against the standards; act as the escalation point for domain-AI design questions
• Contribute to the team’s decisions on standards conformance and its recommendations to the AI Domain Committee

Requirements

• We are hiring for domain modeling judgment and practical AI engineering. Deep expertise in both is rare; strength in one and fluency in the other works.
• Eight or more years building production software, including time at staff, principal, or architect scope where other teams depended on what you owned
• Practical experience designing ontologies, knowledge graphs, or canonical domain models that shipped in production systems
• Hands-on experience with LLM-based or agentic systems: you have built with them and know where they break
• Experience with fine-tuning, RAG, or model evaluation pipelines, and clear judgment about which to reach for
• Hands-on experience with Microsoft Foundry services, including model deployment, agent development, and its evaluation and safety tooling
• Hands-on experience with Azure services more broadly, including the compute, data, identity, and networking building blocks AI workloads run on
• Ability to take a pattern from concept through working proof-of-concept to published standard
• A record of changing technical direction through influence rather than authority, across team boundaries
• Clear writing. The ontology, patterns, and decision records are a real part of the output.

Preferred

• PEO, HCM, payroll, benefits, or adjacent regulated-domain experience
• Working knowledge of ontology and knowledge-representation tooling (OWL, RDF, SHACL, property graphs) or semantic layer design
• Experience with multi-tenant SaaS handling sensitive personal data, including SOC 2 or ISO 27001 environments
• Familiarity with agent frameworks - Microsoft Agent Framework in particular - MCP, and tool-use orchestration
• Knowledge of Snowflake Cortex AI and how it fits alongside a broader AI platform
• Prior AI research experience and formal knowledge-engineering credentials are optional. Domain curiosity is not.