New opportunity

AI Architect

Koch Engineered Solutions · Houston, United States

About this role

Koch Engineered Solutions seeks an AI Architect for its global IT capability. The role partners with IT, Product and business teams to design AI solutions that improve business processes, working across engineering, operations, commercial and finance functions.

Responsibilities

Identify high-value AI and generative AI use cases with Product and business teams, and lead technical solution design and detailed architecture.
Develop reference architectures, patterns, strategy and guardrails aligned with the IT vision, data strategy and responsible-AI standards. Balance complex requirements, project risks and business objectives.
Draw on machine learning, generative AI, data engineering, security, cloud infrastructure and application integration; prototype and evaluate emerging models, agentic frameworks and tools.
Champion model governance, security, data privacy, evaluation and monitoring of production AI solutions.

Required qualifications

Experience designing and deploying AI/ML and generative AI solutions on modern cloud platforms such as Azure AI, Azure OpenAI, AWS Bedrock, Amazon SageMaker, Claude Code or Claude Cowork, or comparable platforms.
Working knowledge of large language models, prompt engineering, retrieval-augmented generation, embeddings and vector stores.
Experience delivering data, analytics or AI/ML solutions in IT; ability to understand customers’ business and operating models and turn them into AI-enabled proposals.
Ability to communicate technical concepts to business and technical audiences at both detailed and high levels.

Preferred qualifications include discrete engineering and manufacturing experience; agentic AI design; deeper knowledge of transformer models, fine-tuning and context engineering; LLMOps/MLOps for deployment, evaluation, version management, monitoring and drift detection; vector databases and retrieval systems; AI orchestration frameworks; responsible-AI governance; AI-supporting data architecture; serverless and event-driven cloud services, microservices and containers; and DevOps practices.

The position can be based in Houston, Texas; Tulsa, Oklahoma; or Wichita, Kansas, with travel as required. It is not eligible for visa sponsorship. Benefits include health coverage, retirement, paid time off and educational assistance, subject to eligibility; some additional benefits may apply.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

AI architectureMachine LearningGenerative AILarge Language ModelsPrompt EngineeringRetrieval-Augmented GenerationEmbeddingsVector StoresAzure AIAzure OpenAIAWS BedrockAmazon SageMakerClaude CodeClaude CoworkData EngineeringCloud InfrastructureApplication IntegrationAI AgentsResponsible AIModel GovernanceAI SecurityData PrivacyModel EvaluationMLOpsLLMOpsFine-TuningVector DatabasesAzure AI SearchpgvectorPinecone

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Select valuable AI use cases
The role partners with Product and business teams to identify high-value AI and generative AI opportunities and turn operating needs into solution proposals. Prepare to explain how you would connect a use case to a business process and decide whether AI is appropriate.
Lead enterprise-aligned architecture
The architect develops detailed designs, reference architectures, patterns and guardrails aligned with the IT vision, data strategy and responsible-AI standards. Prepare to show how a design can meet a specific need while remaining reusable across teams.
Design across technical domains
Solution design spans AI/ML, data engineering, security, cloud infrastructure and application integration; cloud AI deployment experience is a basic qualification. Prepare to explain the interfaces and trade-offs between these domains, not just the model choice.
Evaluate emerging approaches
The role identifies, prototypes and evaluates new models, agentic frameworks and tools for problems new to the organization. Prepare to describe how you would compare an emerging approach with a simpler established option before recommending it.
Build responsible production practices
The role champions model governance, security, privacy, evaluation and ongoing production monitoring. Prepare to distinguish controls needed at design time from evidence and monitoring needed after deployment.

Likely challenges to think through

Reconciling competing requirements
Likely preparation hypothesis: engineering, operations, commercial and finance stakeholders may value different outcomes, while the architect must manage complex requirements and project risk. Prepare a way to make trade-offs explicit and communicate the decision at both business and technical levels.
Moving from prototype to dependable solution
Likely preparation hypothesis: a promising model or agentic prototype may be difficult to integrate and govern in production. Prepare to explain when you would stop experimenting, what evaluation would support the choice, and how privacy and monitoring would affect the architecture.
Balancing local fit with shared patterns
Likely preparation hypothesis: a tailored solution may best fit one workflow but conflict with enterprise architecture, data strategy or guardrails. Prepare to explain when to reuse a reference pattern and how to justify a documented exception.

Practice your answers

Technical practice

How would you identify and prioritize AI use cases with Product and business teams?

What to coverThe role calls for finding high-value use cases and translating business and operating models into solution proposals. Suggested preparation: explain how you would define the process problem, expected value, data needs and risks before recommending AI.

How would you turn an AI use case into a detailed architecture and reusable pattern?

What to coverThe posting calls for detailed solution design and reference architectures aligned with the IT vision, data strategy and responsible-AI standards. Suggested preparation: walk through component boundaries, integration points and the trade-off between a reusable pattern and a use case’s specific needs.

When would you use retrieval-augmented generation rather than prompt changes or fine-tuning?

What to coverRAG, embeddings, vector stores and prompt engineering are core qualifications; fine-tuning is preferred knowledge. Suggested preparation: compare how each approach handles changing information, answer quality and operational effort, including how you would assess retrieval quality.

How would you design and deploy a cloud-based AI solution that integrates with existing applications?

What to coverThe role requires experience deploying AI/ML and generative AI solutions on modern cloud platforms and draws on application integration and security. Suggested preparation: describe the interfaces, data flow, access controls and deployment choices you would evaluate without assuming a particular platform is mandated.

How would you decide whether an action-oriented AI agent is appropriate?

What to coverThe role includes prototyping agentic frameworks, and agent design is listed as preferred experience. Suggested preparation: compare an agent with a simpler workflow, then address tool permissions, human oversight and ways to limit or detect unintended actions.

How would you prototype and evaluate an emerging model or AI tool?

What to coverIdentifying, prototyping and evaluating emerging models and tooling is a stated responsibility. Suggested preparation: define a representative task and baseline, then explain how you would compare quality, risk and practical deployment considerations before recommending adoption.

What guardrails would you put around a production AI solution?

What to coverThe posting explicitly calls for model governance, security, data privacy, evaluation and ongoing monitoring. Suggested preparation: discuss data access, safety checks, ownership of decisions and how findings from production monitoring would lead to changes.

How would you manage an AI solution after deployment?

What to coverLLMOps/MLOps for deployment, evaluation, prompt and version management, monitoring and drift detection are preferred. Suggested preparation: describe what you would version and measure, how you would investigate a quality decline, and when you would roll back or update a component.

What data architecture choices would support an AI solution?

What to coverThe role spans data engineering, while data lakes, knowledge graphs, feature stores and pipelines appear as preferred knowledge. Suggested preparation: explain how the use case would determine data freshness, structure, access and quality needs rather than assuming every solution needs every component.

How would you choose between event-driven services, microservices and containers for an AI application?

What to coverThese cloud application approaches are listed as preferred technical experience. Suggested preparation: compare them against workload behavior, integration needs, operational complexity and the project’s risk profile.

Behavioural practice

Tell me about a time you aligned teams with different goals on a solution design.

What to coverThe role partners with IT, Product and business teams across several functions. Suggested preparation: use a real situation, your actions to understand and reconcile priorities, and the resulting decision or outcome.

Describe a time you balanced complex requirements against project risk.

What to coverThe posting emphasizes understanding a project’s risk profile to choose an appropriate solution. Suggested preparation: give the situation, the trade-offs you surfaced and actions you took, then explain the result.

Give an example of explaining the same technical proposal to business and technical audiences.

What to coverCommunicating AI concepts at both detailed and broad levels is a basic qualification. Suggested preparation: describe the audiences and situation, how you adapted your explanation, and what each audience was able to decide or do afterward.

Tell me about a time you proactively tested an unfamiliar technology to solve a problem.

What to coverThe role calls for experimentation with emerging AI tools and describes a proactive working style. Suggested preparation: cover the problem, the experiment you initiated, how you assessed its limits and the outcome.

Describe a time you gained agreement on a technical guardrail or standard.

What to coverThe architect is expected to lead alignment on architecture patterns and guardrails and champion responsible AI. Suggested preparation: use a real situation, explain how you addressed concerns or resistance, and state the resulting decision or change.

Sharpen your knowledge

Topics to prepare

AI use-case selection and business fit
Revise how to assess value, feasibility, operating-model fit and project risk. Exercise: compare two hypothetical use cases for an engineering or operations team and explain which merits an AI-enabled proposal. Relevant because the role partners with Product and business teams to identify high-value uses and balance business objectives with risk.
Enterprise AI architecture
Revise how data engineering, security, cloud infrastructure and application integration shape an AI solution. Exercise: explain the trade-offs in a reference architecture and how it would align with an IT vision and data strategy. Relevant because the role develops architecture patterns and guardrails across the organization.
LLMs and retrieval-augmented generation
Revise prompt engineering, embeddings, vector stores and the stages of RAG. Exercise: explain when retrieval would help, how source content reaches an answer and where errors can arise. These are named basic-qualification knowledge areas.
Responsible AI and production governance
Revise model governance, security, data privacy, evaluation and monitoring. Exercise: identify risks in a hypothetical production AI workflow and explain corresponding controls and review points. The posting explicitly assigns the architect responsibility for championing these practices.
Agentic AI design
As preferred-qualification preparation, compare conversational and action-oriented agents with a simpler AI workflow. Exercise: explain where an agent might add value, what actions would need guardrails and how you would evaluate the result. Agentic design and evaluation of emerging frameworks are stated in the posting.
AI operationalization
As preferred-qualification preparation, revise LLMOps/MLOps concepts: deployment, prompt and version management, evaluation, monitoring and drift detection. Exercise: describe what changes would trigger a new evaluation or rollback. These practices appear in the preferred qualifications.

Skills to revise

Draft a cloud AI solution design
Practice producing a component diagram and deployment rationale using one of the listed platform families, such as Azure AI/Azure OpenAI or AWS Bedrock/SageMaker. Show data flow, integration and security boundaries. Designing and deploying cloud AI/ML and generative AI solutions is a basic qualification; the exercise is preparation, not a specified interview task.
Build and assess a small RAG prototype
As a preparation exercise, assemble a small permitted document set, generate embeddings, retrieve passages from a vector store and compare answers against source passages. This applies the posting’s required RAG knowledge and its responsibility to prototype and evaluate AI approaches.
Create an architecture decision record
Practice documenting two solution options, assumptions, business value, complex requirements, risks and recommended guardrails. The role leads detailed solution design and must choose an appropriate solution for a project’s risk profile.
Translate a business need for two audiences
Practice turning a hypothetical commercial or finance workflow into an AI-enabled proposal, then present a brief business explanation and a detailed technical explanation. This addresses the stated need to understand operating models and communicate with business and technical audiences.
Define a production evaluation plan
Practice specifying test cases, success criteria, privacy and security checks, monitoring signals and a response to degraded results for a hypothetical AI solution. Evaluation and ongoing production monitoring are explicit responsibilities; prompt versioning and drift detection are preferred-qualification extensions.

Your preparation checklist

Map a business problem to an AI use case
Prepare a one-page hypothetical use case for an engineering, operations, commercial, or finance workflow. State the business objective, stakeholders, expected value, data needs, and reasons AI may or may not be appropriate. This is a preparation exercise, not a stated KES project.
Build a project walkthrough
Select a real AI/ML, generative AI, or data-delivery project you can discuss. Bring a diagram and explain your architecture decisions, cloud deployment, integrations, trade-offs, risks, and your own contributions; do not claim experience you cannot substantiate.
Practice an end-to-end design exercise
Sketch a solution using LLMs, prompts, retrieval-augmented generation, embeddings, and a vector store where justified. Show data flow, access controls, cloud components, application integration, and how the design fits enterprise architecture and data strategy.
Draft a reference pattern and guardrails
Create a short architecture template identifying reusable components, design decision criteria, and review points for security, data privacy, and responsible AI. Be ready to explain how you would balance business objectives against project risk.
Plan evaluation and production operation
Prepare a sample evaluation plan with task-specific quality measures, failure tests, monitoring, version management, and a response to performance drift. Distinguish practices you have implemented from approaches you would propose.
Prepare an emerging-tool comparison
Outline a small prototype plan comparing two model or agentic-tool options on value, reliability, security, and operational effort. Agentic frameworks are a preferred area in the posting, so present them as a choice to evaluate rather than a required solution.
Rehearse stakeholder explanations and questions
Practice a brief business-level proposal and a detailed technical explanation of the same design. Prepare questions about priority use cases, existing IT and data standards, responsible-AI decision-making, and how the role partners with Product and business teams.

Availability last checked: Nov 4, 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in United States. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…