Claude AI Architect (Databricks, Healthcare)

Note:

• 15+ years in IT, with 5+ years as a solution or data architect.
• 6+ years building LLM applications in production, including 4 years+ of hands on work with Claude models and the Claude Developer Platform (Messages API, tool use, structured outputs, prompt caching, thinking, batch processing).

What you will do: Architecture and solution design

• Own end-to-end architecture for Claude-powered applications and agents on Databricks: retrieval-augmented generation, multi-step agents, document intelligence, and workflow automation.
• Target healthcare use cases such as prior authorization review, clinical documentation summarization, claims and appeals processing, member and patient service, care gap closure, risk adjustment, and medical coding support.
• Select and route across the Claude model family (Opus, Sonnet, Haiku) by task, latency, and cost. Design prompt caching, batch processing, and context strategies.
• Decide the deployment path per client: Claude natively in Databricks (Foundation Model APIs, AI Gateway), Claude in Amazon Bedrock, Claude on Google Cloud (Vertex AI), or Claude in Microsoft Foundry.
• Design agent systems with the Claude Developer Platform (Messages API, tool use, structured outputs, thinking, citations, Files API), the Claude Agent SDK, Agent Skills, and Model Context Protocol (MCP) servers that expose Unity Catalog data, Lakebase, and client systems (EHR, claims, CRM) as governed tools.
• Design the Databricks foundation: medallion architecture on Delta Lake; Unity Catalog governance with PHI tagging, row filters, column masks, ABAC, lineage, and audit; Lakeflow (Connect, Spark Declarative Pipelines, Jobs); Mosaic AI Vector Search, Model Serving, AI Gateway, Agent Bricks, and Supervisor Agent; MLflow 3 tracing and evaluation; Databricks Apps.
• Publish reference architectures, architecture decision records, and reusable accelerators for healthcare GenAI on Databricks.

Healthcare compliance and responsible AI

• Design to HIPAA Privacy and Security Rules, HITECH, HITRUST CSF, SOC 2, and 42 CFR Part 2 where applicable: minimum necessary PHI in prompts and logs, de identification (Safe Harbor and Expert Determination), encryption, private networking, audit trails, data residency, zero data retention options, and BAA alignment with every model provider in the chain.
• Define the evaluation and guardrail framework: golden datasets, LLM as judge scorers with clinician review, hallucination and PHI leakage checks, red teaming, human in the loop checkpoints, model risk documentation, and production monitoring.
• Architect ingestion and semantic layers for healthcare standards: HL7 v2, FHIR R4, C-CDA, X12 EDI (837, 835, 270/271, 278), and terminologies (ICD-10, CPT, HCPCS, SNOMED CT, LOINC, RxNorm, NDC).

Leadership and client engagement

• Lead technical discovery, estimation, and roadmaps. Present architecture and tradeoffs to client CIO, CDO, CMIO, CISO, and compliance stakeholders.
• Set coding, prompt, and evaluation standards. Mentor and review Claude AI engineers and Databricks data engineers, onshore and offshore.
• Own solution FinOps: token spend, DBU consumption, serving endpoint sizing, and chargeback models.

Must have

Claude and generative AI

• 6+ years building LLM applications in production, including 4 years+ of hands on work with Claude models and the Claude Developer Platform (Messages API, tool use, structured outputs, prompt caching, thinking, batch processing).
• Experience designing agentic systems with the Claude Agent SDK and MCP, including custom MCP servers, tool permissioning, and multi-agent patterns.
• Proven RAG and agent evaluation practice: retrieval quality, groundedness, task success, latency, and cost per transaction.
• Working knowledge of Claude deployment paths: natively in Databricks (Foundation Model APIs, AI Gateway), Amazon Bedrock, Google Cloud (Vertex AI), and Microsoft Foundry, including BAA and data retention terms for each.

Databricks (deep, hands-on)

• 5+ years on Databricks, including 2+ years architecting lakehouse platforms at enterprise scale.
• Expert in Delta Lake, Unity Catalog (fine grained access, ABAC, lineage, audit, Delta Sharing, Clean Rooms), Lakeflow (Connect, Spark Declarative Pipelines, Jobs), Databricks SQL and Photon, serverless compute, and Databricks Asset Bundles for CI/CD.
• Expert in Mosaic AI: Model Serving (Foundation Model APIs, external models, provisioned throughput), AI Gateway, Vector Search, Agent Framework, Agent Bricks, MLflow 3 (tracing, evaluation, prompt registry, model registry), Databricks Apps, Lakebase, and AI/BI Genie.
• Strong PySpark, Spark SQL, and Python, with a record of performance tuning and cost optimization on Databricks.
• Cloud architecture on Azure or AWS: identity (Entra ID or IAM), networking (Private Link, VNet or VPC injection), key management, secret scopes, service principals, and SCIM provisioning.

Healthcare domain

• 5+ years delivering data or AI solutions for payers, providers, PBMs, or life sciences.
• Deep working knowledge of HIPAA, HITECH, and HITRUST as applied to AI systems, including de identification methods, minimum necessary rules, and BAA requirements.
• Hands on with healthcare data standards: HL7 v2, FHIR R4, C-CDA, X12 EDI (837, 835, 270/271, 278), claims and encounter data, EHR data (Epic Clarity or Caboodle, Oracle Health), and clinical terminologies.
• At least two healthcare AI use cases taken to production, for example prior authorization, clinical summarization, claims automation, member services, risk adjustment, or HEDIS and Stars quality.

Architecture and leadership

• 15+ years in IT, with 5+ years as a solution or data architect.
• Track record of taking GenAI solutions from pilot to production with measurable business outcomes.
• Clear communication with executive, clinical, and compliance audiences. Experience leading onshore and offshore teams.
• Able to work US business hours. [Work authorization requirement, if any.]

Nice to have

• Claude for Healthcare connectors and Agent Skills (CMS Coverage Database, ICD-10, NPI Registry) and Claude for Enterprise deployments under a BAA.
• Databricks healthcare accelerators: FHIR ingestion with dbignite, patient 360, claims analytics; Databricks Clean Rooms for multi party data collaboration.
• Clinical NLP and medical coding automation; vision models for scanned forms, faxes, and handwritten documents (prior auth packets, referrals, lab requisitions).
• Multi agent orchestration with Supervisor Agent, or LangGraph and OpenAI Agents SDK agents deployed on Databricks Apps.
• EHR and payer integration: Epic (Clarity, Caboodle, FHIR APIs), Oracle Health, Availity, or clearinghouse connections.
• Awareness of FDA software as a medical device guidance and ONC HTI-1 transparency rules for predictive decision support.
• Life sciences: GxP and 21 CFR Part 11 validated environments.
• Terraform for Databricks and cloud infrastructure; Microsoft Fabric or Snowflake alongside Databricks.
• Databricks partner or Champion program involvement; conference talks or published reference architectures.

Education and certifications

• Bachelor's degree in Computer Science, Engineering, or a related field; Master's preferred.
• Databricks Certified Generative AI Engineer Associate, held or obtained within 90 days of start.
• Databricks Certified Data Engineer Professional or Machine Learning Professional preferred.
• Cloud architect certification preferred: Microsoft Certified Azure Solutions Architect Expert or AWS Certified Solutions Architect Professional.
• Healthcare credentials a plus: HL7 FHIR Proficiency, CPHIMS, or AHIMA CHPS.
• Anthropic's Claude developer training a plus.