Enterprise Architect

*Description*
The AI Enterprise Architect is responsible for defining, evolving, and driving the enterprise architecture for artificial intelligence across the client's environment.
This is not a theoretical, advisory-only, or documentation-focused architecture role. The AI Enterprise Architect must be a hands-on technology leader who can move rapidly from an ambiguous business problem to an executable architecture, working reference implementation, production deployment, and measurable business outcome.
The role operates at the intersection of enterprise architecture, AI engineering, data, security, cloud platforms, integration, product delivery, and business strategy. This individual will establish the enterprise direction for generative AI, agentic AI, machine learning, intelligent automation, AI-enabled applications, and shared AI platform capabilities.
The AI Enterprise Architect will work within a Fortune 500 environment while supporting an AI program that operates with the urgency, experimentation, adaptability, and delivery expectations of a startup. The successful candidate must be comfortable making architecture decisions in a rapidly evolving technology landscape, challenging conventional approaches, eliminating unnecessary complexity, and personally driving initiatives through organizational and technical barriers.
This individual must be capable of seeing the entire enterprise AI ecosystem while remaining close enough to implementation to validate architectures, examine code and configurations, build prototypes, identify delivery risks, and distinguish production-ready capabilities from demonstrations and vendor claims.
Key Responsibilities
Enterprise AI Strategy and Target Architecture
* Define and maintain the enterprise AI target architecture, transition architectures, capability model, platform strategy, and multiyear architecture roadmap.
* Translate business strategy and operating priorities into executable AI capabilities, architecture investments, and delivery sequences.
* Establish the architectural direction for generative AI, agentic AI, machine learning, intelligent automation, AI-assisted decision-making, and AI-enabled business processes.
* Define clear boundaries and relationships between enterprise AI platforms, domain solutions, shared services, data platforms, enterprise applications, and external AI providers.
* Ensure that project-level AI decisions support enterprise scalability, interoperability, security, reuse, and long-term maintainability.
* Identify opportunities to consolidate overlapping technologies, eliminate duplicated capabilities, and prevent uncontrolled AI platform and vendor sprawl.
* Develop architecture options and recommendations that explicitly address business value, delivery speed, cost, risk, technical debt, vendor dependency, and operational complexity.
* Maintain a current enterprise view of AI capabilities, platforms, models, agents, integrations, data dependencies, vendors, risks, and strategic initiatives.
Hands-On AI Architecture and Delivery
* Lead AI initiatives from problem definition and architecture through implementation, production deployment, adoption, and measurable outcomes.
* Develop working prototypes and reference implementations to validate architecture decisions, platform capabilities, integration approaches, security controls, and delivery feasibility.
* Review source code, prompts, agent definitions, tool configurations, retrieval pipelines, model configurations, APIs, infrastructure, and deployment pipelines as needed to validate solution quality.
* Work directly with engineering teams to resolve architecture and implementation issues rather than limiting involvement to reviews or recommendations.
* Rapidly diagnose delivery blockers, simplify overengineered approaches, reduce unnecessary scope, and establish practical paths to production.
* Define production-readiness criteria and ensure that AI solutions meet requirements for reliability, security, performance, observability, supportability, cost, and business continuity.
* Distinguish clearly between proof of concept, pilot, minimum viable product, production capability, and enterprise platform.
* Remain personally accountable for architecture outcomes, not only architecture artifacts or review completion.
Generative and Agentic AI Architecture
* Design enterprise-grade architectures for large language models, multimodal models, AI assistants, autonomous and semi-autonomous agents, and AI-enabled applications.
* Define patterns for single-agent and multi-agent orchestration, tool use, planning, reasoning, memory, state management, delegation, and human approval.
* Establish architecture standards for retrieval-augmented generation, structured retrieval, knowledge graphs, semantic search, and enterprise knowledge access.
* Define patterns for context engineering, prompt management, structured outputs, model routing, fallback, caching, and workload segmentation.
* Architect secure agent access to enterprise systems, APIs, data, workflows, and external services.
* Define patterns for Model Context Protocol, agent-to-agent communication, enterprise APIs, event-driven interactions, and tool integration.
* Establish controls around nondeterministic model behavior, including deterministic validation, approval checkpoints, execution boundaries, and exception handling.
* Evaluate when AI agents are appropriate and when conventional software, workflow automation, rules engines, APIs, or analytics provide a better solution.
* Prevent the use of generative AI or agents where the architecture introduces unnecessary cost, risk, latency, or operational complexity.
Enterprise AI Platform Architecture
* Define the architecture for shared enterprise AI platform capabilities, including model access, model gateways, agent runtime services, retrieval services, evaluation services, security controls, observability, and cost management.
* Establish reusable AI services, platform components, reference architectures, templates, development patterns, and deployment patterns.
* Define enterprise model access, model selection, model portability, workload routing, quota management, and vendor abstraction strategies.
* Design workload, tenant, domain, environment, and data isolation patterns appropriate to enterprise risk and operating requirements.
* Establish architectural standards for proprietary, open-weight, hosted, and internally operated models.
* Define integration patterns between AI platforms and enterprise cloud, data, identity, security, integration, application, and observability platforms.
* Partner with platform engineering, cloud infrastructure, data, cybersecurity, and application teams to establish a scalable AI operating environment.
* Ensure that platform capabilities are implemented as usable products and services rather than architecture concepts that delivery teams cannot practically adopt.
Data and Knowledge Architecture
* Define data and knowledge architecture required to support AI models, agents, applications, evaluation, analytics, and business processes.
* Establish patterns for structured, semi-structured, and unstructured data access.
* Define architectures using relational, document, graph, vector, search, streaming, and analytical technologies based on workload requirements.
* Establish standards for embeddings, chunking, indexing, metadata, reranking, retrieval, source attribution, and information freshness.
* Define approaches for enterprise taxonomies, ontologies, semantic models, knowledge graphs, and reusable domain knowledge.
* Ensure appropriate data lineage, provenance, ownership, quality, classification, access control, retention, and usage restrictions.
* Define requirements for training, fine-tuning, inference, retrieval, evaluation, monitoring, and feedback datasets.
* Ensure that AI responses and actions can be traced to authoritative enterprise information where required.
* Identify situations where weak data, fragmented ownership, or poor knowledge management must be corrected rather than hidden behind an AI interface.
Integration and Distributed Systems Architecture
* Define and enforce AI integration patterns across enterprise applications, cloud platforms, SaaS products, data platforms, APIs, workflows, and external services.
* Architect synchronous and asynchronous APIs, event-driven interactions, messaging, streaming, workflow orchestration, and long-running business processes.
* Establish standards for identity propagation, delegated authorization, agent identity, workload identity, and service-to-service authentication.
* Define system-of-record ownership, transaction boundaries, data contracts, API contracts, semantic contracts, and integration responsibilities.
* Ensure AI solutions appropriately address retries, timeouts, idempotency, circuit breakers, error handling, compensating transactions, and recovery.
* Define patterns for human-in-the-loop workflows, approvals, exception handling, and escalation.
* Maintain the enterprise AI integration map, documenting dependencies and touchpoints among AI capabilities and enterprise platforms.
* Identify reusable enterprise services and integrations that can accelerate multiple AI initiatives.
AI Evaluation and Quality Engineering
* Establish enterprise standards for evaluating AI models, agents, retrieval systems, prompts, workflows, and business outcomes.
* Define offline and online evaluation approaches appropriate to each AI use case.
* Establish task-specific quality measures covering accuracy, groundedness, relevance, completeness, factuality, safety, latency, reliability, and cost.
* Define evaluation methods for retrieval quality, agent trajectories, tool selection, tool execution, workflow completion, and human intervention.
* Establish regression testing for prompts, models, retrieval configurations, agents, workflows, integrations, and platform changes.
* Define adversarial testing, red-team testing, edge-case testing, and failure-mode testing requirements.
* Establish human evaluation, adjudication, feedback, and quality review processes where automated evaluation is insufficient.
* Define quality gates for development, testing, release, production rollout, and model or configuration changes.
* Ensure that AI quality is measured continuously and not determined solely through demonstrations, subjective user feedback, or initial pilot results.
AI Operations, LLMOps, MLOps, and AgentOps
* Define the enterprise operating model for AI solution development, testing, deployment, monitoring, support, and retirement.
* Establish standards for CI/CD, infrastructure as code, configuration management, prompt management, model versioning, agent versioning, and environment promotion.
* Define release strategies, including feature flags, canary deployments, controlled rollout, rollback, and fallback.
* Establish observability requirements across models, agents, prompts, retrieval, tools, APIs, workflows, infrastructure, and business outcomes.
* Define standards for distributed tracing, token usage, latency, errors, tool calls, retrieval results, quality, safety events, and consumption costs.
* Establish controls for quotas, rate limits, capacity, concurrency, resource utilization, and unbounded consumption.
* Define operational responsibilities, service-level expectations, support models, incident response, escalation, recovery, and post-incident review.
* Establish FinOps practices for model inference, AI platform consumption, data movement, storage, and supporting infrastructure.
* Ensure AI capabilities can be operated reliably by enterprise support and engineering teams after initial delivery.
AI Security and Risk Architecture
* Define security architecture and threat-modeling requirements for AI models, agents, platforms, applications, data, tools, integrations, and workflows.
* Establish controls for direct and indirect prompt injection, sensitive information disclosure, insecure outputs, model manipulation, data poisoning, and supply-chain risk.
* Define controls to prevent excessive agency, overprivileged access, unauthorized actions, and uncontrolled execution.
* Establish least-privilege access patterns for models, agents, tools, APIs, data, and enterprise systems.
* Define requirements for agent identity, workload identity, secrets management, credential handling, sandboxing, isolation, and egress control.
* Establish human authorization requirements for consequential, irreversible, financial, customer-facing, security-sensitive, or legally significant actions.
* Define monitoring and response requirements for misuse, anomalous behavior, model extraction, data leakage, abuse, and unexpected consumption.
* Partner with cybersecurity, privacy, legal, compliance, and risk teams to ensure AI controls are technically implementable and operationally effective.
* Ensure that security requirements are built into architecture and delivery rather than added after implementation.
Responsible AI and Governance
* Translate AI policies, principles, legal requirements, and risk expectations into specific architecture and engineering controls.
* Define risk-based architecture requirements based on use-case impact, data sensitivity, autonomy, audience, and potential consequences.
* Establish architecture checkpoints, approval requirements, exception processes, and escalation paths without unnecessarily slowing delivery.
* Define requirements for transparency, explainability, disclosure, human oversight, traceability, audit evidence, and accountability.
* Maintain or contribute to the enterprise inventory of AI use cases, models, agents, platforms, vendors, risks, and accountable owners.
* Establish lifecycle requirements covering