Director, Enterprise AI Architect

Description:

The Enterprise AI Architect is responsible for defining and governing the enterprise AI platform architecture, ensuring secure, scalable, compliant, and cost-effective AI solutions across the organization. This role provides strategic technical leadership for AI capabilities spanning data integration, retrieval-augmented generation (RAG), model orchestration, agentic AI, observability, FinOps, and security. Serving as the organization’s AI architecture authority, the position establishes enterprise standards, reviews vendor solutions, and drives alignment across Commercial, R&D, Regulatory, TechOps, Data, and Information Security teams. The role partners with senior business and technology leaders to enable AI innovation, ensure responsible AI adoption, and develop a sustainable enterprise AI ecosystem that supports long-term business objectives.

Essential Functions:

• AI platform architecture — end-to-end ownership of AIP architecture across the integration layer, RAG pipelines, orchestration, and model layer; ensure consistency across all four layers and vendor-built solutions

• Data-to-AI contract and integration layer design — define how Gold datasets are exposed to AI, chunking/embedding/indexing standards, and RBAC propagation; own ingestion pipeline standards and drive the Phase 1 to Phase 2 transition

• FinOps and observability — own model cost tracking, token usage monitoring, and AI observability tooling; establish baseline measurement and ongoing cost governance

• Vendor technical submission review — review every vendor AI solution design for RBAC implementation, prompt injection exposure, data leakage controls, audit trail completeness, and platform standards compliance

• Model gateway design and agent development toolkit — define model routing/fallback architecture; establish standards and governance for agentic AI development

• AI security review collaboration — partner with Global InfoSec to define and execute the technical AI security review process

Governance, standards, and business alignment — define metadata standards, AI design patterns, and model usage policies; partner with Commercial, R&D, Regulatory, and TechOps

Additional Responsibilities:

• Guide third-party implementation vendors on architecture decisions and integration patterns; prevent vendor lock-in; ensure effective knowledge transfer so Amneal retains architectural ownership

• Define citizen development standards in partnership with the Enterprise Data Architect

• Work with AI Solutions Engineers embedded in business functions

Education:

• Bachelors Degree (BA/BS) Computer Science, Data Engineering, Information Systems, or related field — or equivalent work experience - Required
• Master Degree (MS/MA) Computer Science, Data Science, or related field - Preferred

Experience:

• 12 years or more in IT/enterprise architecture experience, including enterprise-level AI platform architecture (not project/solution level)

Specialized Knowledge:

• Four-layer enterprise AI architecture (data platform, integration, orchestration, model layer)
• Model gateway / routing / fallback design
• RAG pipeline design at production scale
• AI governance frameworks and Architecture Review Board (ARB) standards
• FinOps for AI workloads - Model cost monitoring, token usage tracking, cost allocation across a multi-use-case AI portfolio, built from scratch
• Data-to-AI integration design - RBAC propagation, metadata standards, data contracts from governed data to AI systems
• Multi-LLM provider architecture - OpenAI, Anthropic, AWS Bedrock, or equivalent; vendor-agnostic by design
• AWS data/AI ecosystem - S3, Glue, SageMaker, Bedrock, or equivalent
• Databricks/lakehouse platform - Databricks or similar
• Regulated industry deployment - Pharma, financial services, or healthcare; audit trail, access control, compliance constraints in practice
• Vendor governance - Directing external vendors on architecture decisions and knowledge transfer
• 21 CFR Part 11 / GxP AI deployment
• Agentic AI architecture - Multi-agent orchestration, memory management, tool use, human-in-the-loop design
• AI security - IAM for AI workloads, DLP for LLM I/O, prompt controls, audit logging, AI threat modeling
• Citizen development governance - Guardrails for non-technical users building AI workflows (M365 Copilot Studio, Claude Projects, or equivalent)
• Agent observability tooling - Braintrust, Lang Smith, or …