Agentic AI Data Architect
RemoteUnited States
Job Summary
Define end-to-end architecture for agentic AI-enabled platforms across data, AI, orchestration, and integration layers, including hands-on POCs. Design and govern multi-step production workflows, RAG, vector search, and MCP tool access patterns. Establish integration frameworks connecting lakehouses, ODS, document systems, underwriting systems, and third-party APIs using Python/.NET hybrid microservices. Determine where to apply GenAI versus deterministic logic, design multimodal integrations, and create prompt lifecycle strategies with evaluation and optimization. Implement AI safety guardrails, PII masking, hallucination control, and ModelOps/PromptOps frameworks to ensure explainability and auditability. Lead technical execution across AI, data, and platform teams, guiding engineers and driving stakeholder communication. Requires 10–15+ years in software/data/AI engineering with 4–6+ years in GenAI architecture.
Required Qualifications
- 10–15+ years in software/data/AI engineering
- 4–6+ years in AI/ML/GenAI architecture
- Strong experience in designing enterprise-scale platforms and distributed systems
- Bachelor's or Master's in Computer Science, Engineering, Data Science, or related field
- Hands-on architect with ability to balance strategy + execution
- Define end-to-end architecture for agentic AI-enabled platform across data, AI, orchestration, and integration layers
- Design and govern agentic orchestration framework for multi-step production workflows
- Establish architecture patterns for - RAG and grounding, Vector search and retrieval, MCP tool access layer, prompt management and evaluation
- Have a deep understanding of Agentic coding and best practices of using Agentic coding for large scale implementations
- Familiarity in implementing A2A or similar frameworks in a large scale environment
- Define integration architecture across - Lakehouse, ODS, document systems, Underwriting systems and third-party APIs
- Design configurable, metadata-driven framework for multi-LOB onboarding
- Define API/microservices patterns (Python/.NET hybrid)
- Define where and how to use - GenAI vs deterministic logic, agentic workflows vs pipeline workflows
- Establish multimodal integration approach combining structured, unstructured, and external data
- Design prompt lifecycle, evaluation, and optimization strategy
- Define AI safety and guardrails (PII, hallucination control, policy constraints)
- Establish ModelOps and PromptOps frameworks
- Ensure explainability, auditability, and traceability of AI outputs
- Lead technical execution across AI, data, and platform teams
- Guide engineers (AI, data, full-stack) and ensure alignment with architecture
- Drive technical decisions and stakeholder communication
- GenAI & Agentic Frameworks - Semantic Kernel/ LangGraph (or similar orchestration frameworks)
- LLM integration (Azure OpenAI, OpenAI APIs, etc.)
- Prompt engineering, prompt lifecycle design
- Retrieval & RAG - Azure AI Search (indexing, vector search, hybrid search)
- Embedding pipelines and retrieval optimization
- RAG design, grounding strategies, context management
- Tool Access & Integration - MCP (Model Context Protocol) architecture and tool design
- API design (FastAPI / REST / microservices)
- Integration with enterprise systems and third-party APIs
- AI Safety & Governance - NVIDIA NeMo Guardrails
- Microsoft Presidio (PII detection/masking)
- Guardrails for prompt injection, hallucination control
- Evaluation & ModelOps - Azure AI Foundry (model hosting, versioning, monitoring)
- Evaluation frameworks (LLM-as-judge, test datasets)
- Prompt/version control, cost/latency monitoring
- DevOps & Observability - CI/CD pipelines (Azure DevOps / GitHub Actions)
- Logging, monitoring, observability (App Insights, etc.)
- Performance tuning and scalability
Desired Qualifications
- Domain (good to have): Insurance / reinsurance / financial services
Hiring someone like this?
Get your role in front of qualified candidates on Sorce.