AI Solution Architect

Lead AI discovery and use case prioritization, scoring opportunities on data sensitivity, cost at scale, latency, feasibility, and governance exposure. Design end-to-end architectures spanning RAG, agentic workflows, data pipelines, model serving, and guardrails. Make model-selection recommendations across closed (GPT, Claude, Gemini) and open-weight (Llama, Mistral, Qwen) options, applying a structured hard-attribute / soft-attribute framework. Choose the deployment target deliberately: Azure, AWS, on-prem NVIDIA AI factory, or hybrid, and document the rationale. Define interface specifications between AgreeYa’s application layer and partner-owned infrastructure (for example, model-serving endpoint contracts, performance baselines), protecting against handoff and dependency risk. Own the application-level governance design: NIST AI RMF alignment, risk tiering, human-in-the-loop placement, audit and explainability requirements. Set delivery standards and review the work of AI Engineers and MLOps Engineers for architectural soundness.

Demonstrated production AI/ML solution architecture, not only pilots and proofs of concept. Deep RAG fluency: chunking strategy, embedding model selection, vector search, retrieval evaluation. Working knowledge of agentic patterns, orchestration (LangChain / LangGraph), and tool integration (including MCP). Strong grasp of model selection, fine-tuning vs RAG trade-offs, and inference cost/latency economics. Able to lead technical client conversations and defend design decisions to a skeptical technical audience.

Must be able to architect and reason fluently across all three of the following, and recommend between them: Azure AI: Azure AI Foundry, Azure OpenAI Service, Azure AI Search, Azure ML. AWS AI: Amazon Bedrock, SageMaker, OpenSearch, Lambda-based serving. On-prem NVIDIA AI factory: NVIDIA AI Enterprise (NVAIE), NIM microservices, Triton Inference Server, NeMo and NeMo Guardrails, Run:ai, TensorRT-LLM, and quantized/air-gapped deployment (GGUF, vLLM).