AI Architect

Role Summary

The AI Architect owns the end-to-end technical design across all four workstreams and is the primary technical counterpart to Client's IT leadership, security team, and Executive Sponsor. This person sets the reference architecture for the Enterprise Hub, the RAG assistant, and the AI-DLC toolchain, then defends those designs in security review. They are accountable for the engagement being architecturally coherent — one set of patterns  can reuse, not four disconnected pilots.

Key Responsibilities

• Lead WS#1 discovery: facilitate use case identification workshops, apply the value/complexity/risk/data-readiness/regulatory scoring rubric, build the data source inventory, and produce a prioritized backlog with a signed-off pilot use case and at least three Phase 2 candidates.

• Design the target architecture for the Generative AI Enterprise Hub — Model Gateway spanning Amazon Bedrock, Anthropic, Cohere, and custom SageMaker endpoints; centralized guardrails; prompt management; agentic orchestration accelerators; usage and cost telemetry with allocation tags.

• Design the RAG reference architecture on the AWS-maintained GenAI Chatbot on AWS solution: ingestion and chunking strategy, vector store selection, retrieval and re-ranking approach, citation/grounding design, and document-level authorization model via Amazon Cognito.

• Own the security architecture: least-privilege IAM, KMS and encryption boundaries, VPC and endpoint design, Bedrock Guardrails policy set (content filtering, PII redaction, responsible AI enforcement), and identity federation via AWS IAM Identity Center or agency SSO.

• Lead the STRIDE threat model and AI security review, and present both to the security team as a formal acceptance criterion.

• Define the AI Operating Model for WS#3: decision rights, quality gates, human-in-the-loop checkpoints, and the guardrails governing how AI-assisted development operates inside agency policy.

• Establish evaluation methodology and quality thresholds for retrieval and generation (groundedness, citation accuracy, answer relevance, refusal behavior), and the acceptance evidence tied to each milestone.

• Direct the technical work of the two AI Engineers and the DevOps Engineer; review designs and code at the pattern level.

• Present at executive readouts, produce the post-engagement roadmap, and lead knowledge transfer to  architects.

Required Qualifications

• 8+ years in software/solutions architecture, with 2+ years designing and shipping generative AI systems in production.

• Deep hands-on AWS: Bedrock, SageMaker, Lambda, API Gateway, AppSync, Step Functions, OpenSearch/Aurora pgvector or equivalent vector stores, Cognito, IAM Identity Center, KMS, CloudWatch.

• Demonstrated end-to-end RAG architecture experience — not just prototypes. Must be able to speak precisely about chunking trade-offs, hybrid retrieval, re-ranking, and why a given system hallucinates.

• Threat modeling experience (STRIDE or equivalent) and the ability to lead a security review with a skeptical government security team.

• Regulated-data architecture experience: HIPAA, PHI/PII handling, data classification, audit and logging requirements.

• Excellent written and verbal communication. This role writes the documents that get signed.

Preferred Qualifications

• AWS Certified Solutions Architect – Professional; AWS Certified Machine Learning – Specialty.

• Prior delivery in Medicaid, MMIS, health and human services, or another state agency environment.

• Experience with Kiro, MCP server architecture, Amazon Q Business/Developer, and QuickSight Q.

• Experience standing up an internal AI platform or model gateway serving multiple application teams.

• FinOps literacy — cost attribution, throttling, and rate-limiting strategy for shared model access.