Teach AI what our business actually means.
PEO and HCM are dense domains. Co-employment, payroll, benefits, workers' compensation, and multi-state compliance carry meaning that a general-purpose model simply does not have. Those domains anchor a family of PrismHR platforms, and your job is to encode that meaning so every AI capability we publish reasons about our domain correctly.
You will join the AI Domain team, which owns the enterprise standards, reference architectures, approved model catalog, and governance safeguards for AI across the company. Within it 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.
Four areas of ownership. The canonical ontology as a semantic layer spanning all PrismHR platforms — entities, relationships, business rules, and terminology across co-employment, payroll, benefits administration, workers' compensation, onboarding, and compliance — built with the product teams who own each data model, and governed as a living artifact. Agent workflow modeling: business processes decomposed into agentic workflows with tool boundaries, decision points, escalation paths, and human-in-the-loop checkpoints, defining where agents may act autonomously and where they must defer, especially around payroll, money movement, and compliance. Fine-tuning strategy: when to fine-tune, when to retrieve, when prompting is enough, with dataset curation standards covering labeling, provenance, retention, and residency, and domain-specific evaluation harnesses that measure accuracy in PEO and HCM rather than on generic benchmarks. And the intelligence pattern library: retrieval strategies, reasoning templates, agent scaffolds, and validation guards, each proven by a working proof-of-concept before publication and documented with its failure modes.
You will engage with product teams from design through go-live, advise on use-case feasibility and risk, act as the escalation point for domain-AI design questions, and contribute to standards conformance decisions and recommendations to the AI Domain Committee.
You will need:
- 8+ 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
- 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 Microsoft Foundry: model deployment, agent development, and its evaluation and safety tooling
- Hands-on Azure more broadly — 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
We are hiring for two things: domain modeling judgment and practical AI engineering. Deep expertise in both is rare. Strength in one and fluency in the other works.
Preferred: PEO, HCM, payroll, benefits, or adjacent regulated-domain experience · ontology and knowledge-representation tooling (OWL, RDF, SHACL, property graphs) or semantic layer design · multi-tenant SaaS handling sensitive personal data, including SOC 2 or ISO 27001 environments · agent frameworks, Microsoft Agent Framework in particular, plus MCP and tool-use orchestration · 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.
The schedule: 6:00 PM – 3:00 AM IST, Monday to Friday, with the first four hours from the Noida office and the rest from home. For a role that sets enterprise standards and advises product teams across the company, full overlap is what makes the influence real rather than nominal. Your mornings and afternoons are yours.
Competitive compensation aligned to architect scope, health coverage, conference and learning support, and ownership of the layer every AI capability in the company depends on. Apply now.
We may use artificial intelligence (AI) tools to support parts of the hiring process, such as reviewing applications, analyzing resumes, or assessing responses and identifying potential inconsistencies or verification signals in application materials based on available information. These tools assist our recruitment team but do not replace human judgment. Final hiring decisions are ultimately made by humans. If you would like more information about how your data is processed, please contact us.
