Ai Rag Jobs in Oak Brook, IL (NOW HIRING)

Top Skills' Details

1. AI solution architecture across business, application, data, integration, security, and infrastructure domains
2. Generative AI, RAG, and Agentic AI Architecture (Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), Agentic AI and multi-agent systems, Prompt engineering and context orchestration, Vector databases, embeddings, reranking, and retrieval strategies, Model selection, routing, fallback, and optimization.
3. AI Security, Identity, and Governance

*Must Have previous experience as an Architect with A.I Enterprise Experience*

*Must have

For the AI Solution Architect role,

• Generative AI
• RAG / knowledge architecture
• Agent architecture
• AI evaluation
• AI security
• LLMOps / AgentOps and production operations
• Model architecture
• Enterprise integration
• AI economics

Description

The AI Solutions Architect serves as the technical architecture leader for enterprise AI solutions within the AI Center of Excellence and reports directly to the Vice President of Enterprise Architecture. This role is responsible for designing secure, scalable, supportable, and economically sustainable AI-enabled solutions from initial concept through production deployment and ongoing operation.

The AI Solutions Architect owns the end-to-end architecture of assigned AI initiatives, including business requirements, artificial intelligence models, agents, data and knowledge sources, integrations, identity, security, infrastructure, observability, operational support, and governance controls. The role ensures that AI capabilities are not developed as isolated experiments, but as enterprise-grade solutions that integrate with existing business processes, platforms, applications, and data ecosystems.

The architect works directly with business stakeholders, product owners, engineering teams, Cybersecurity, Data and Analytics, Infrastructure, Legal, Risk, and vendor partners while ensuring alignment with enterprise architecture strategy, standards, governance, and technology roadmaps. The individual must be able to translate business objectives into actionable architecture, validate technical designs through hands-on analysis and prototyping, and clearly communicate architectural decisions, risks, costs, and trade-offs. This position requires demonstrated experience delivering production generative AI, retrieval-augmented generation, machine learning, and agentic AI solutions. Experience limited to strategy, presentations, vendor demonstrations, or proofs of concept does not satisfy the requirements of the role.

AI Solution Architecture

• Design end-to-end architectures for enterprise AI solutions, including generative AI, retrieval-augmented generation, conversational AI, predictive machine learning, intelligent automation, computer vision, speech, and agentic AI capabilities.

• Translate business requirements into comprehensive technical solution designs covering applications, models, agents, data, integrations, security, cloud infrastructure, observability, operations, and governance.

• Determine whether artificial intelligence is appropriate for a given business problem and recommend alternative technical approaches when AI does not provide sufficient value, reliability, or economic benefit.

• Define current-state, target-state, and transitional architectures for AI initiatives, including technical dependencies, shared capabilities, implementation phases, and architecture risks.

• Ensure AI solutions align with enterprise architecture standards, cloud strategies, approved technology platforms, cybersecurity requirements, data governance policies, and operational support models.

• Create architecture acceptance criteria and validate that proposed solutions meet functional, technical, security, operational, and business requirements before production deployment.

Generative AI and Model Architecture

• Design production-grade generative AI solutions using commercial, open-source, hosted, dedicated, and privately deployed models.

• Evaluate and select language, vision, speech, embedding, reranking, and multimodal models based on solution quality, latency, cost, context requirements, data sensitivity, deployment options, scalability, supportability, and vendor risk.

• Design multi-model architectures that support model routing, fallback, portability, workload specialization, and reduced dependency on a single model provider.

• Define prompt architecture, context assembly, structured-output requirements, response validation, model fallback, caching, rate limiting, and error-handling patterns.

• Evaluate model performance using repeatable technical and business criteria rather than vendor benchmarks or demonstration results alone.

• Maintain awareness of model capabilities, limitations, licensing considerations, deployment constraints, and rapidly changing AI platform capabilities.

Agentic AI Architecture

• Design secure and reliable AI agents that can reason, use tools, maintain state, interact with enterprise applications, and execute controlled business workflows.

• Define agent responsibilities, tool boundaries, memory models, workflow states, delegation rules, approval requirements, and termination conditions.

• Design single-agent and multi-agent solutions using deterministic workflow controls around nondeterministic model behavior.

• Establish architecture patterns for human-in-the-loop review, escalation, exception handling, retries, timeouts, circuit breakers, compensating actions, and emergency termination.

• Define secure agent-to-user, agent-to-agent, and agent-to-tool interaction patterns.

• Prevent uncontrolled agent autonomy by enforcing least privilege, constrained tool access, transaction limits, validation rules, and approval gates for consequential actions.

• Partner with AI engineering teams to establish consistent agent development, orchestration, testing, deployment, and lifecycle management standards.

Retrieval-Augmented Generation and Knowledge Architecture

• Design enterprise retrieval-augmented generation solutions across structured, unstructured, document, transactional, graph, and operational data sources.

• Define document ingestion, parsing, chunking, metadata enrichment, embedding, indexing, retrieval, reranking, citation, and knowledge-refresh strategies.

• Design lexical, semantic, vector, hybrid, graph-enhanced, and structured retrieval patterns based on the characteristics of each use case.

• Ensure retrieval solutions preserve source-system security, authorization, data classification, retention, and user entitlements.

• Establish patterns for authorization-aware retrieval, source attribution, content freshness, provenance, and deletion.

• Define controls for retrieval poisoning, outdated content, duplicate information, conflicting sources, inappropriate data exposure, and unsupported model responses.

• Evaluate retrieval quality, answer relevance, groundedness, citation accuracy, and knowledge coverage before production deployment.

AI Evaluation and Quality Engineering

• Define measurable quality standards and evaluation strategies for generative AI, retrieval, machine learning, and agentic AI solutions.

• Establish golden datasets, benchmark scenarios, regression suites, adversarial tests, and business acceptance criteria.

• Define evaluation methods for accuracy, relevance, groundedness, hallucination, toxicity, bias, safety, retrieval quality, tool selection, tool-call accuracy, agent trajectory, and task completion.

• Implement automated evaluation gates within AI development and deployment pipelines.

• Define the appropriate use of human evaluation, expert review, LLM-based evaluation, deterministic testing, and statistical analysis.

• Ensure model, prompt, retrieval, agent, and tool changes are tested against previous production behavior before release.

• Establish production quality thresholds, monitoring requirements, rollback criteria, and exception-management processes.

AI Security, Identity, and Trust Architecture

• Design AI solutions in accordance with enterprise cybersecurity, privacy, identity, compliance, and risk-management requirements.

• Perform AI-specific threat modeling covering prompt injection, indirect prompt injection, data poisoning, retrieval poisoning, sensitive-data exposure, model extraction, system-prompt leakage, insecure tool invocation, excessive agency, and downstream code execution.

• Define identity propagation and authorization patterns across users, agents, models, tools, APIs, applications, data sources, and external services.

• Design least-privilege access, workload identities, delegated authorization, service accounts, session isolation, tenant isolation, and approval controls.

• Ensure agents cannot access data, tools, or transactions beyond the permissions of the requesting user or approved system identity.

• Define security controls for AI gateways, model endpoints, vector stores, knowledge bases, MCP servers, external tools, plugins, third-party models, and vendor services.

• Establish auditability that records who initiated an AI request, what context was used, which decisions were made, which tools were invoked, who approved an action, and what action was executed.

• Partner with Cybersecurity teams to conduct red-team exercises, abuse-case testing, vulnerability assessments, and production-readiness reviews.

Data and Integration Architecture

• Design data flows and integration architectures connecting AI solutions with enterprise applications, cloud services, data platforms, customer platforms, contact-center systems, operational systems, and external providers.

• Define integration patterns using REST, GraphQL, gRPC, APIs, messaging, event streaming, batch processing, change-data capture, and workflow orchestration.

• Define tool and function-calling contracts, schema validation, idempotency, rate limiting, error handling, transaction boundaries, and compensating actions.

• Design integrations with enterprise platforms such as Salesforce, Snowflake, ServiceNow, ERP solutions, digital platforms, and internal business applications.

• Develop and maintain solution-level integration and data-flow documentation identifying all systems, interfaces, ownership boundaries, security controls, and dependencies.

• Ensure data contracts, metadata, lineage, data quality, classification, retention, privacy, consent, masking, and deletion requirements are incorporated into solution designs.

• Identify shared services, reusable connectors, common APIs, enterprise tools, and platform capabilities that can reduce duplication and accelerate delivery.

• Prevent AI agents and applications from becoming uncontrolled or redundant integration layers.

Cloud and AI Platform Architecture

• Design AI workloads using approved enterprise cloud platforms, infrastructure services, network patterns, and deployment standards.

• Architect model endpoints, AI gateways, agent runtimes, vector and graph stores, knowledge services, containerized workloads, serverless components, and Kubernetes-based deployments.

• Define private connectivity, secrets management, encryption, key management, workload isolation, network segmentation, and access controls.

• Design for scalability, high availability, regional resilience, recoverability, capacity management, and service continuity.

• Determine when to use managed AI services, vendor-hosted capabilities, open-source technologies, dedicated deployments, or internally operated platforms.

• Define infrastructure-as-code, environment management, deployment automation, configuration management, and platform support requirements.

• Work with Infrastructure and Platform Engineering teams to ensure AI workloads are production-ready, monitored, supportable, and aligned with enterprise cloud standards.

LLMOps, MLOps, and AgentOps

• Define lifecycle-management practices for models, prompts, agents, tools, datasets, embeddings, knowledge indexes, evaluation suites, and configuration artifacts.

• Establish versioning, traceability, approval, deployment, promotion, rollback, and retirement requirements across development, testing, staging, and production environments.

• Design CI/CD pipelines that include automated testing, security scanning, evaluation gates, policy checks, and production-readiness validation.

• Define canary, shadow, blue-green, phased-release, and feature-flag patterns for AI solution deployment.

• Establish model, prompt, agent, retrieval, and tool rollback mechanisms, kill switches, and emergency disablement procedures.

• Define experiment tracking, release documentation, environment reproducibility, and audit evidence requirements.

• Ensure production incidents can be traced to the specific model, prompt, agent, tool, data, retrieval index, code, and configuration versions involved.

AI Observability and Production Operations

• Define end-to-end observability for user requests, prompt construction, context assembly, retrieval, model calls, agent decisions, tool executions, workflow transitions, human approvals, responses, and downstream actions.

• Establish logging, tracing, monitoring, alerting, dashboards, and service-level objectives for AI solutions.

• Define operational metrics for latency, availability, model usage, token consumption, cost, failure rates, retrieval quality, groundedness, safety violations, tool accuracy, task completion, agent loops, escalation rates, and user outcomes.

• Ensure observability integrates with enterprise monitoring, logging, incident-management, and support platforms.

• Define production support models, ownership boundaries, runbooks, escalation processes, incident response, problem management, and recovery procedures.

• Establish controls for detecting model regressions, data drift, prompt failures, retrieval degradation, unexpected agent behavior, and cost anomalies.

• Partner with engineering and operations teams to ensure AI solutions can be supported outside of the origina