Agentic AI Architect

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)

Perform hands on development and 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

Define AI safety and guardrails (PII, hallucination control, policy constraints)
Establish ModelOps and PromptOps frameworks
Ensure explainability, auditability, and traceability of AI outputs

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 workflows
Establish architecture patterns for - RAG and grounding, Vector search and retrieval, MCP tool access layer, prompt management and evaluation

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

Experience: 10–15+ years in software/data/AI engineering with 4–6+ years in AI/ML/GenAI architecture
Background: Strong experience in designing enterprise-scale platforms and distributed systems
Domain (good to have): Insurance / reinsurance / financial services
Education: Bachelor's or Master's in Computer Science, Engineering, Data Science, or related field
Profile Type: Hands-on architect with ability to balance strategy + execution

Technical skills:

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

Originally posted on Himalayas

Desbloquea con Pro