New opportunity

Applied AI Engineer (Hybrid)

RTX · Farmington, United States of America

About this role

RTX Enterprise Services seeks an Applied AI Engineer to design, build, evaluate and deploy production-grade AI and machine learning solutions for business and engineering problems across the company. The role takes solutions from prototypes to secure, scalable production deployments and measurable outcomes, working with business, architecture, platform, data, product and cybersecurity teams.

Responsibilities

Build AI and ML solutions using traditional machine learning, generative AI, retrieval-augmented generation and agentic approaches. Develop agents and multi-step workflows that use tools, enterprise applications and data with appropriate human oversight.
Engineer retrieval and grounding systems using enterprise knowledge, embeddings, search, prompts, context and memory. Integrate solutions with APIs, data sources and applications, including through Model Context Protocol where appropriate.
Develop tested software, APIs, integrations and reusable AI components. Select models based on quality, reliability, latency, cost, security, scalability and business needs; create systematic evaluations and use traces, telemetry, feedback and failure analysis to improve performance and reliability.

Required qualifications

A university degree in computer science, AI, machine learning, engineering or a related STEM discipline with at least 8 years of relevant professional experience; alternatively, an advanced degree in a related field with at least 5 years of relevant professional experience.
At least 3 years of hands-on AI/ML development, integration or deployment experience, including moving capabilities beyond experimentation into production or production-like environments.
Python and production-quality software engineering experience; experience with generative AI and LLM applications, external-system integration, machine learning fundamentals, model evaluation, source control, automated testing, CI/CD, containerization and production deployment.

Preferred qualifications include production agents or multi-agent systems; retrieval, vector databases or knowledge graphs; platforms such as LangGraph, CrewAI, IBM watsonx, AWS Bedrock or n8n; agent interoperability, AI evaluation and observability, MLOps, cloud-native deployment, AI security and Responsible AI. Cross-functional collaboration and clear communication are also preferred.

This full-time hybrid role requires residence within commuting distance of one of the eight listed U.S. locations; onsite time is determined with the employee’s leader. Applicants must be U.S. persons; a security clearance is not required. The stated salary range is 107,500–204,500 USD. Eligible hires may receive healthcare, retirement, leave and other benefits, with eligibility dependent on the business unit and position.

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

Skills for this role

PythonMachine LearningGenerative AILarge Language ModelsAI AgentsMulti-agent SystemsRetrieval-Augmented GenerationPrompt EngineeringContext EngineeringEmbeddingsVector DatabasesEnterprise SearchKnowledge GraphsModel Context ProtocolAPI IntegrationModel EvaluationAI ObservabilityMLOpsCI/CDAutomated TestingContainerizationDockerKubernetesLangGraphCrewAIIBM watsonxAWS Bedrockn8nCloud ServicesResponsible AI

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

Production-ready delivery
The role takes AI/ML solutions from prototypes to secure, scalable deployments and measurable business outcomes. It calls for tested, maintainable Python software, APIs, CI/CD, containerization, and reusable components.
Grounded enterprise AI
The posting calls for retrieval and context engineering over enterprise knowledge, using approaches such as embeddings, search, prompts, and memory to support reliable outputs.
Agents with controlled integrations
A core duty is building multi-step agents that use tools, enterprise applications, APIs, and data with appropriate human oversight. Model Context Protocol is mentioned as an option where appropriate.
Evidence-driven model and system choices
The engineer must compare approaches on quality, reliability, latency, cost, security, scalability, and business needs; develop systematic evaluations; and use traces, telemetry, feedback, and failure analysis to improve performance.
Cross-functional production rollout
The role partners with business, architecture, platform, data, evaluation, and cybersecurity teams to move solutions into secure production environments.

Likely challenges to think through

Balancing agent autonomy and oversight
Preparation hypothesis, not a confirmed RTX problem: multi-step tool use may improve task completion while increasing the risk of unintended actions. Prepare to explain how you would define tool permissions, approval points, and ways to recover from failed steps.
Measuring groundedness against operating constraints
Preparation hypothesis, not a confirmed RTX problem: stronger retrieval or a different model may improve answer quality but add latency and cost. Prepare a comparison approach with representative evaluation cases, failure categories, and explicit selection criteria.
Moving integrations across enterprise environments
Preparation hypothesis, not a confirmed RTX problem: connecting data and applications while meeting security and deployment needs can complicate the move from prototype to production. Prepare to discuss interface design, testing, observability, and coordination with platform and cybersecurity teams; the posting also identifies cloud, hybrid, on-premises, and restricted environments as areas of learning.

Practice your answers

Technical practice

How would you choose between traditional ML, RAG, and an agentic workflow for an enterprise problem?

What to coverThe posting calls for selecting appropriate AI and ML approaches for business and engineering problems. Suggested answer focus (inference): compare task structure, available data, need for grounded answers or external actions, operational complexity, and measurable outcomes.

How would you design retrieval and grounding over enterprise knowledge sources?

What to coverThe role includes embeddings, vector and enterprise search, prompts, memory, and other grounding techniques. Suggested answer focus (inference): explain source selection, access-aware retrieval, context construction, and how you would test whether answers are supported by retrieved material.

How would you engineer a multi-step AI agent with tool use and human oversight?

What to coverThe posting describes agents that use tools, enterprise applications, data, and appropriate human oversight. Suggested answer focus (inference): cover workflow state, tool boundaries, error recovery, and when a person should review or approve an action.

How would you integrate an AI solution with enterprise APIs and data sources?

What to coverThe role calls for integrations using standard interfaces and, where appropriate, Model Context Protocol. Suggested answer focus (inference): discuss interface contracts, authentication and permissions, handling external-system failures, and when MCP offers an advantage over a direct API integration.

How would you select a model for a production AI application?

What to coverThe posting names quality, reliability, latency, cost, security, scalability, and business requirements as selection factors. Suggested answer focus (inference): describe a comparative test using representative tasks and explain how you would resolve conflicting metrics.

How would you build systematic evaluations for an LLM application?

What to coverThe role calls for evaluation cases that measure solution performance, including groundedness and task completion. Suggested answer focus (inference): define representative and failure-focused cases, scoring criteria, a baseline, and checks for regressions before release.

How would you diagnose an AI system that performs well in tests but fails in production?

What to coverThe posting calls for using traces, telemetry, user feedback, and failure analysis to improve reliability. Suggested answer focus (inference): explain how you would isolate retrieval, model, prompt, tool, and integration failures, then verify a fix with evaluations and monitoring.

What would production-quality Python delivery look like for an AI service?

What to coverThe required experience includes tested, maintainable Python, source control, CI/CD, containerization, and production deployment. Suggested answer focus (inference): cover API and component boundaries, automated tests, reproducible builds, deployment checks, and safe changes to an existing service.

How would you secure an AI application that handles enterprise data and tool actions?

What to coverThe posting emphasizes secure deployment, cybersecurity collaboration, and preferred familiarity with AI security, privacy, governance, and Responsible AI. Suggested answer focus (inference): discuss least-privilege access, data handling, controls on tool actions, and how you would validate those controls with relevant partners.

How would you move a traditional ML solution into production and keep it useful?

What to coverThe role includes traditional ML, model evaluation, production deployment, and preferred experience with model lifecycle management and MLOps. Suggested answer focus (inference): cover a suitable baseline, validation against the business objective, serving considerations, and monitoring for performance changes.

Behavioural practice

Tell me about a time you moved an AI or ML capability beyond a prototype.

What to coverThe posting emphasizes taking solutions through production or production-like deployment to measurable outcomes. Suggested answer structure (inference): give the situation, your engineering actions and cross-team dependencies, and the observed result without overstating impact.

Describe a time you resolved a difficult AI system failure.

What to coverThe role calls for diagnosing behavior and improving robustness and reliability. Suggested answer structure (inference): explain the failure's context, how you investigated and tested a remedy, and what changed afterward.

Tell me about a time you aligned technical and nontechnical teams on an AI solution.

What to coverThe posting names collaboration with business, architecture, platform, data, product, evaluation, and cybersecurity teams. Suggested answer structure (inference): describe the differing needs, how you communicated decisions and trade-offs, and the outcome.

Describe a time you had to make a difficult quality, latency, cost, or security trade-off.

What to coverThe role explicitly requires evaluating those competing factors against business needs. Suggested answer structure (inference): state the constraints, evidence behind your decision, who you consulted, and the result.

Tell me about a time you independently tackled an ambiguous technical problem.

What to coverThe preferred qualifications include independently solving complex problems and communicating technical trade-offs. Suggested answer structure (inference): explain how you clarified the goal, tested options, sought input where needed, and measured the result.

Sharpen your knowledge

Topics to prepare

Model selection and ML fundamentals
The posting requires ML fundamentals and selecting approaches against quality, reliability, latency, cost, security, scalability and business needs. As preparation, compare traditional ML, an LLM and a retrieval-based approach for a hypothetical business task; state what evidence would justify each choice.
Retrieval and grounding design
The role includes enterprise-data retrieval, embeddings, search, prompts, context and memory. As preparation, diagram how a question reaches relevant source material and becomes a grounded answer; identify where irrelevant retrieval or unsupported claims could arise.
Agent workflow boundaries
The role calls for multi-step agents that use tools and enterprise applications with human oversight. As preparation, sketch a workflow and mark its tool permissions, retained state, failure paths and points requiring human review.
Evaluation and production feedback
The posting calls for systematic evaluation and improvement using traces, telemetry, user feedback and failure analysis. As preparation, define test cases and measures for groundedness, task completion and reliability, then explain how a failed case would guide a change.
Secure production architecture
The role moves prototypes into secure, scalable deployments with architecture, platform, data and cybersecurity teams. As preparation, draw a deployment boundary diagram for a hypothetical AI service and explain data access, integration and operational tradeoffs; the diagram is a preparation exercise, not a stated RTX design.

Skills to revise

Build and test a Python AI service
The posting requires Python and production-quality, tested, maintainable software and APIs. As preparation, implement a small Python API around a model call with input validation, structured output and automated tests for errors and expected responses.
Implement a grounded retrieval flow
The role involves retrieval-augmented generation and enterprise knowledge sources. As preparation, use sample documents to build a retrieval-and-answer flow, test an unanswerable question, and inspect whether the response is supported by retrieved material.
Integrate a tool-using workflow
The role includes external-system integration and agents that use tools with human oversight; MCP is mentioned as an interoperability approach. As preparation, connect a small multi-step workflow to a mock API, handle a failed tool response and require approval before a consequential action.
Run evaluations and diagnose failures
The posting requires model evaluation and calls for traces, telemetry and failure analysis. As preparation, run a repeatable set of success, unsupported-answer and tool-failure cases; record outcomes and traces, then document one targeted improvement and its measured effect.
Package a deployment-ready change
Source control, automated testing, CI/CD, containerization and production deployment are required experience. As preparation, containerize the small service, make its tests run in a CI/CD pipeline, and rehearse a versioned change and rollback plan.

Your preparation checklist

Build a requirements evidence sheet
Map available examples to the posted degree and experience paths, at least three years of hands-on AI/ML work, Python, LLM applications, integrations, and production or production-like deployment. Note the U.S.-person and commuting-distance conditions without assuming eligibility.
Prepare a production project walkthrough
Choose a real AI/ML project and outline the problem, your decisions, collaborators, prototype-to-deployment path, failures, and measurable outcomes where available. The role emphasizes reliable production systems and business results.
Sketch an enterprise AI design
As a practice exercise, diagram a grounded retrieval or agent workflow: data and API access, search and context, tool use, human oversight, and security boundaries. Explain when a simpler ML or non-agent approach would be preferable; MCP is an option in the posting, not a mandate.
Create an evaluation scorecard
Draft example test cases and a comparison of candidate approaches on quality, groundedness, task completion, reliability, latency, cost, security, and scale. Show how traces, telemetry, user feedback, and failure analysis would guide improvements.
Collect software delivery evidence
Prepare a concise account of Python code, API or external-system integration, source control, automated tests, CI/CD, containerization, and deployment from work you can discuss. Identify what was reusable and how it was maintained.
Rehearse a cross-functional decision story
Outline one example of communicating an AI tradeoff with business, architecture, platform, data, product, or cybersecurity partners. State each team's concern, the decision made, and the result without implying you worked with every group.
Write targeted employer questions
Ask which business or engineering problems this role would address first, how teams assess production AI quality and security, what deployment environments are relevant, and how hybrid onsite time is set for the position.

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 of America. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…