New opportunity

Senior AI Engineer - Data & MLOps (Teradyne, Remote)

Teradyne · North Reading, United States

About this role

Teradyne seeks a remote Senior AI Engineer to build ML and AI solutions for its IT organization and establish the MLOps/LLMOps practices needed to operate them reliably. The role reports to the Enterprise AI/Data Product Manager within Enterprise Architecture and Data and bridges hands-on solution development with production ownership.

Responsibilities

Translate IT and business problems into predictive and classification models, generative AI assistants, RAG workflows, agents, and intelligent automation. Design, train, evaluate, deploy, and integrate solutions with enterprise data, secure APIs, and MCP servers.
Move internally developed solutions into production under release-readiness standards. Define model versioning, registry, retraining, promotion, deprecation, drift, and degradation criteria; provide technical input for operations runbooks and handoffs.
Build reusable, multi-environment MLOps/LLMOps pipelines, CI/CD, automated offline and online evaluation, regression tests, and infrastructure- and configuration-as-code for models, prompts, and agents. Make deployments observable, auditable, and supportable.
Establish technical standards and playbooks; mentor engineers. Implement responsible AI guardrails, security controls against risks including prompt injection and unauthorized data access, logging, audit trails, and monitoring of quality, reliability, performance, and cost.

Requirements

8–10 years in ML/AI or software engineering, including at least 3 years building and operating production ML/AI systems, plus recent hands-on generative and agentic AI experience. Bachelor’s or advanced degree in Computer Science, Data Science, Engineering, or a related field.
Strong Python and SQL; experience with scikit-learn, PyTorch, or TensorFlow and with one or more listed enterprise AI platforms: Azure AI Foundry, Microsoft Copilot Studio, Anthropic Claude, Cursor, Google Vertex AI, or Snowflake Cortex AI.
Experience with MLOps/LLMOps, Azure DevOps or GitHub Actions, Azure Machine Learning, MLflow, model lifecycle management, agentic orchestration, RAG, vector databases, embeddings, prompt engineering, and MCP/API integration. Knowledge of evaluation, observability, responsible AI, security, and governance; multicloud experience with Azure as primary and Google Cloud as secondary, integrating Snowflake and Microsoft Fabric data. Mentoring, collaboration, communication, and analytical skills are required.

Compensation and benefits: Base salary is $133,900–$223,900, varying with experience, skills, demand, and location. The role is eligible for discretionary bonuses. Benefits include health coverage, retirement savings plans, paid vacation and holidays, insurance, and tuition assistance.

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

Skills for this role

PythonSQLscikit-learnPyTorchTensorFlowAzure AI FoundryMicrosoft Copilot StudioAnthropic ClaudeCursorGoogle Vertex AISnowflake Cortex AIAzure DevOpsGitHub ActionsAzure Machine LearningMLflowMicrosoft AzureGoogle CloudSnowflakeMicrosoft FabricMLOpsLLMOpsCI/CDModel lifecycle managementModel evaluationDrift detectionRetrieval-augmented generationVector databasesEmbeddingsPrompt engineeringAI agents/agentic AI orchestration frameworks?

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

Deliver useful AI solutions for IT
Posting: Translate IT and business problems into predictive models, assistants, RAG workflows, agents, and automation; evaluate them for measurable impact. Preparation: Be ready to explain how you select an approach and measure whether it solves the underlying problem.
Make releases production-ready
Posting: Define release readiness, model versioning, registry, retraining, promotion, deprecation, and drift criteria for internally developed solutions. Preparation: Outline a lifecycle from evaluation through monitoring and remediation.
Build repeatable MLOps and LLMOps
Posting: Create reusable, multi-environment pipelines, CI/CD, configuration-as-code, and offline and online regression evaluation for models, prompts, and agents. Preparation: Describe how you would make releases testable, auditable, and supportable.
Protect and observe deployed AI
Posting: Apply responsible-AI guardrails and controls against prompt injection and unauthorized data access; log and monitor quality, reliability, performance, and cost. Preparation: Connect safeguards to concrete failure modes and monitoring signals.
Prepare the team for ownership
Posting: Establish standards, reference patterns, and playbooks; mentor engineers and provide technical input for operations handoffs. Preparation: Explain how documentation and reusable patterns help others operate a solution after launch.

Likely challenges to think through

Choosing the right solution complexity
Possible challenge (inference): The remit spans classical ML and generative or agentic AI, so a more capable-looking approach may add cost and operational risk without improving the outcome. Prepare a comparison using task fit, evaluation results, reliability, and measurable impact.
Handoff without losing technical accountability
Possible challenge (inference): This engineer defines readiness and remediation criteria while the AI Platform Operations Team owns runbooks and executes scheduled retraining and registry operations. Prepare to explain how clear thresholds, dependencies, failure modes, and ownership boundaries support a workable handoff.
Grounding AI securely across systems
Possible challenge (inference): Combining enterprise data, retrieval, agents, MCP servers, and APIs across the listed platforms can create tension between access, usefulness, and security. Prepare an example design that limits data access and tests both answer quality and unauthorized-access risks.

Practice your answers

Technical practice

How would you choose between a predictive model, a classification model, a GenAI assistant, and automation for an IT problem?

What to coverSuggested answer focus: explain how you would define the decision or task, establish a baseline, and choose an approach based on data, error costs, integration needs, and measurable business impact. Describe how you would test whether the added complexity of an AI solution is justified.

How would you develop and validate a predictive or classification model using enterprise data?

What to coverSuggested answer focus: cover SQL-based data checks, leakage prevention, training and validation design, model selection, and metrics aligned to the IT problem. Explain how you would use Python and an appropriate ML framework to make experiments reproducible.

How would you build a RAG workflow grounded in trusted enterprise data?

What to coverSuggested answer focus: discuss ingestion, embeddings, retrieval quality, access controls, and how you would check that answers are supported by retrieved material. The posting names Snowflake and Microsoft Fabric data and an Azure-primary, Google-Cloud-secondary environment; explain integration choices without assuming a particular architecture.

How would you integrate an AI agent with enterprise APIs and MCP servers?

What to coverSuggested answer focus: describe tool permissions, input and output validation, failure handling, and safeguards for actions that affect enterprise systems. Explain when an agent is warranted instead of a simpler assistant or workflow.

How would you evaluate a generative AI solution before and after release?

What to coverSuggested answer focus: propose representative offline test cases and regression checks for answer quality, grounding, safety, and task completion. Explain what online signals would reveal regressions or gaps that offline tests missed.

How would you design a reusable, multi-environment MLOps or LLMOps pipeline?

What to coverSuggested answer focus: walk through tracked experiments, automated tests, promotion gates, deployment, and rollback using the posting’s examples of Azure DevOps or GitHub Actions, Azure Machine Learning, and MLflow where relevant. Address how infrastructure and configuration as code would make models, prompts, and agents auditable.

How would you manage model versions, drift, retraining, and deprecation?

What to coverSuggested answer focus: distinguish data drift from degraded task performance, then define evidence-based thresholds and review points for tuning, retraining, promotion, or retirement. Explain how registry records and evaluation results would support those decisions.

What would you include in a production-readiness handoff for an internally developed AI solution?

What to coverSuggested answer focus: cover tested behavior, dependencies, failure modes, monitoring, rollback, and clear operating criteria. The posting says release readiness follows AI Platform Operations Team standards and that an operations partner owns runbooks and on-call processes; explain the technical input you would provide rather than assuming you author those processes.

How would you mitigate prompt injection and unauthorized data access in a RAG or agent workflow?

What to coverSuggested answer focus: discuss permission-aware retrieval, constrained tool access, treatment of retrieved content as untrusted, and tests for attempted misuse. Explain how logging and audit trails can aid investigation while respecting data-handling requirements.

How would you diagnose declining quality or rising cost in a deployed AI solution?

What to coverSuggested answer focus: identify the quality, reliability, latency, and cost signals you would inspect and how you would trace a change across data, retrieval, model or prompt versions, and integrations. Compare possible remedies against their operational and business impact.

Behavioural practice

Tell me about a time you translated an ambiguous business or IT problem into a measurable AI solution.

What to coverSuggested answer focus: use situation, action, and result to show how you clarified the need, aligned with stakeholders, and chose success measures. State the actual outcome without implying that every problem required AI.

Describe a time you handed a solution from development into production ownership.

What to coverSuggested answer focus: use situation, action, and result to explain how you worked with operations partners on readiness, dependencies, failure modes, and support criteria. Include what changed as a result of their feedback.

Tell me about a time you helped engineers adopt a technical standard or reusable practice.

What to coverSuggested answer focus: use situation, action, and result to describe the capability gap, how you mentored or documented a pattern, and how you assessed adoption. Keep the example grounded in your actual contribution.

Describe a time a production ML or AI solution degraded or behaved unexpectedly.

What to coverSuggested answer focus: use situation, action, and result to show how you investigated signals, coordinated remediation, and communicated the impact. Explain any monitoring or release changes made afterward.

Tell me about a time you raised a security, data-access, or responsible-AI concern during delivery.

What to coverSuggested answer focus: use situation, action, and result to explain the risk, how you discussed trade-offs with others, and what safeguard or decision followed. Avoid claiming a compliance outcome unless your example supports it.

Sharpen your knowledge

Topics to prepare

Choosing ML and AI approaches
Suggested revision exercise: For an IT request, compare a predictive model, a classifier, a GenAI assistant, and intelligent automation; state how you would select an approach and measure its impact. The posting calls for translating IT and business problems into fit-for-purpose solutions.
Production lifecycle and ownership
Suggested revision exercise: Map a model from experiment through registry, promotion, monitoring, retraining, and deprecation; define release-readiness and handoff criteria. The role owns the transition of internally developed solutions to production under AI Platform Operations Team standards.
RAG and agent architecture
Suggested revision exercise: Diagram how an agent would use embeddings, retrieval, enterprise data, an MCP server, and a secure API; identify where grounding and access checks belong. The posting explicitly covers RAG, agentic orchestration, and enterprise integration.
Responsible AI and operational risk
Suggested revision exercise: Threat-model a RAG assistant for prompt injection and unauthorized data access, then specify guardrails, audit evidence, and quality, reliability, and cost measures. These controls and monitoring duties are explicit in the posting.
Enterprise AI stack and data boundaries
Suggested revision exercise: Sketch the responsibilities and data flow in a deployment using Azure as the primary cloud, Google Cloud as secondary, and enterprise data from Snowflake or Microsoft Fabric. The role spans these environments and asks for hands-on experience with one or more listed AI platforms, not every platform.
Standards, playbooks, and mentoring
Suggested revision exercise: Outline a short development-to-support playbook and how you would teach another engineer to use it. The role establishes reusable standards and mentors engineers while preparing the team to own solutions.

Skills to revise

Python and SQL solution development
Suggested revision exercise: Write Python and SQL for a small predictive or classification workflow, including data preparation, model training, and an evaluation result using one listed ML framework, such as scikit-learn. Strong Python and SQL and ML-framework experience are stated requirements.
Multi-environment pipeline implementation
Suggested revision exercise: Draft a CI/CD pipeline in Azure DevOps or GitHub Actions that tracks an experiment with MLflow, runs tests, and gates promotion across environments; specify what would be versioned. The posting calls for reusable MLOps/LLMOps pipelines, Azure Machine Learning, MLflow, and auditable releases.
RAG and agent integration
Suggested revision exercise: Prototype or diagram a retrieval pipeline with embeddings and a vector database, then specify an agent's calls through an MCP server or secure API and test denied access. This reflects the stated hands-on RAG, agent, and integration work.
Automated evaluation and regression testing
Suggested revision exercise: Define an offline test set and online evaluation measures for one predictive model and one GenAI workflow; write pass/fail criteria for a release. The posting explicitly requires automated offline and online evaluation and regression tests.
Monitoring and production handoff
Suggested revision exercise: Draft a monitoring checklist with drift or degradation thresholds, retraining triggers, logs, dependencies, failure modes, and escalation information. The role defines remediation criteria and supplies technical input for operations runbooks and handoffs.

Your preparation checklist

Map the stated requirements to evidence
The posting asks for production ML/AI experience, recent generative and agentic AI work, Python, SQL, and relevant platforms. Prepare a one-page evidence map with specific examples, dates, your contributions, and any gaps to discuss honestly.
Prepare a production solution walkthrough
The role spans predictive or classification models through deployment. Prepare one project walkthrough covering the problem, data, model selection, evaluation, integration, measurable outcome, and what you owned after release.
Sketch an enterprise RAG or agent design
The posting calls for retrieval, embeddings, agents, secure APIs, and MCP integration. As a practice exercise, draw a hypothetical IT assistant architecture showing data access, orchestration, evaluation, and failure handling; do not assume access to Teradyne data.
Show how you automate releases
The role specifies multi-environment pipelines, CI/CD, Azure Machine Learning, MLflow, and testing for models, prompts, and agents. Prepare a pipeline diagram or sanitized artifact showing versioning, automated checks, promotion, and rollback.
Draft a production ownership handoff
The posting distinguishes this role’s technical input from the AI Platform Operations Team’s runbooks and operations. Prepare a sample release-readiness checklist with dependencies, drift and degradation thresholds, retraining criteria, failure modes, and handoff notes.
Explain evaluation and safeguards
The role calls for offline and online evaluation, regression tests, observability, audit trails, and controls for prompt injection and unauthorized access. Prepare a compact test matrix with metrics, monitoring signals, security tests, and responses to failures.
Prepare a mentoring example
The posting expects technical standards, reusable patterns, and engineer mentoring. Bring one concrete example of a playbook or standard you introduced, how you taught it, and what changed; use a proposed approach if you lack a direct example.
Bring questions about operating boundaries
Ask how the AI Enablement Team, this engineer, and the AI Platform Operations Team divide release approval and ongoing ownership; which IT use cases are priorities; and how success is measured across Azure, Google Cloud, Snowflake, and Microsoft Fabric.

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…