New opportunity

Staff, Data Science & Applied AI

Warner Bros. Discovery · Atlanta, United States of America

About this role

As a senior individual contributor on Warner Bros. Discovery’s Enterprise Data & AI Solutions team, the Staff, Data Science & Applied AI role develops scalable analytics, production-grade models and data products for businesses spanning Studios, Streaming, Linear Networks, Consumer Products, Games and Direct-to-Consumer platforms. The role partners with Product, Engineering and business stakeholders to turn complex challenges into solutions with measurable enterprise impact.

Responsibilities

Design, deploy and monitor statistical, predictive and machine learning models for customer analytics, forecasting, personalization, optimization and content performance. Apply techniques including ensembles, gradient boosting, deep learning, NLP, time-series forecasting and recommendations.
Build LLM-based solutions for knowledge retrieval, content intelligence, metadata enrichment, summarization and workflow automation. Develop prompts, evaluation frameworks and guardrails; architect RAG pipelines across structured and unstructured enterprise data; and adapt foundation models using techniques such as LoRA or adapters when appropriate.
Evaluate production AI systems for hallucinations, bias, output quality, latency and cost. Work with Responsible AI and Governance teams on enterprise policies, privacy and ethical practices.
Collaborate with Data Engineering and DevOps to productionize solutions in scalable cloud environments. Develop CI/CD and model-lifecycle processes covering experiment tracking, versioning and automated retraining; monitor model and prompt drift, performance degradation and data integrity. Build reusable frameworks and automation to reduce manual work.

Qualifications

Bachelor’s degree, MS or higher in Computer/Data Science, Engineering, Mathematics, Statistics or a related quantitative discipline; 8+ years of relevant data science experience and 2+ years of GenAI experience. A track record of production AI/ML solutions with measurable business impact is required.
Hands-on experience with LLM deployment, prompt optimization, RAG, vector databases, embeddings, semantic search, parameter-efficient model adaptation and LLM evaluation. Experience with enterprise AI platforms such as AWS Bedrock, Azure OpenAI, Databricks Model Serving or Snowflake Cortex, plus understanding of Responsible AI, privacy and model governance, is sought. Exposure to multimodal AI and agent-based workflows is a plus.

The position is listed in Atlanta, Georgia, United States of America, as full-time at 40 hours per week.

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

Skills for this role

Data ScienceStatistical ModelingPredictive ModelingMachine LearningDeep LearningNatural Language ProcessingTime Series ForecastingRecommendation SystemsEnsemble MethodsGradient BoostingGenerative AILarge Language ModelsPrompt EngineeringFew-Shot LearningStructured Output DesignRetrieval-Augmented GenerationVector DatabasesEmbedding ModelsSemantic SearchPineconeWeaviateOpenSearchFAISSModel Fine-TuningLoRAInstruction TuningLLM EvaluationHallucination DetectionResponsible AIData Privacy and Governance in AI Systems and Data Products Applied in Practice

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 measurable enterprise impact
Source-backed: This senior individual-contributor role translates business challenges into production-grade models and data products across WBD businesses, working with Product, Engineering and business stakeholders. Preparation suggestion: Be ready to explain how a solution’s success was defined, measured and communicated.
Build robust predictive analytics
Source-backed: The role develops models for customer analytics, forecasting, personalization, optimization and content performance, with rigorous validation and performance tracking. Preparation suggestion: Review how you choose methods and validation designs for different business decisions.
Develop useful LLM and RAG applications
Source-backed: Duties include knowledge retrieval, content intelligence, metadata enrichment, summarization and automation; RAG must integrate structured and unstructured enterprise data. Preparation suggestion: Be ready to explain retrieval, prompt and model-adaptation choices against a specific use case.
Evaluate AI quality and govern risk
Source-backed: The role measures hallucinations, bias, output quality, latency and cost, and works with Responsible AI and Governance teams on privacy and enterprise policies. Preparation suggestion: Prepare to describe evaluation criteria, guardrails and what would trigger a change or rollback.
Make solutions maintainable in production
Source-backed: The role partners with Data Engineering and DevOps on cloud deployment, CI/CD, experiment tracking, versioning, retraining and drift monitoring; it also builds reusable frameworks. Preparation suggestion: Review how you would assign operational ownership and reduce manual work across teams.

Likely challenges to think through

Reconciling different business objectives
Preparation hypothesis, not a confirmed WBD problem: A solution spanning Studios, Streaming or other listed businesses may face different definitions of value, making reuse harder than a single-use-case build. Prepare an example of agreeing on metrics with stakeholders while keeping the technical design adaptable.
Balancing retrieval coverage with trustworthy answers
Preparation hypothesis, not a confirmed WBD problem: Combining structured and unstructured enterprise data in RAG may improve coverage while increasing irrelevant retrievals or unsupported answers. Prepare to discuss how you would test retrieval separately from generation and use evidence-based guardrails.
Balancing release speed with ongoing reliability
Preparation hypothesis, not a confirmed WBD problem: Frequent model or prompt changes can improve results but complicate versioning, drift detection and reproducibility. Prepare a rollout approach with baselines, monitoring thresholds and rollback criteria.
Trading model quality against operational constraints
Preparation hypothesis, not a confirmed WBD problem: A stronger LLM solution may cost more or respond more slowly, while privacy and bias controls can constrain design choices. Prepare to explain how you would compare options using quality, latency, cost and governance measures together.

Practice your answers

Technical practice

How would you choose and validate a model for a content-performance or customer-analytics problem?

What to coverSuggested preparation: Compare a statistical baseline with approaches named in the posting, such as gradient boosting or deep learning, based on the data and decision being supported. Explain how you would prevent leakage, validate robustness, and connect performance measures to business value.

How would you design and evaluate a forecasting solution across changing business conditions?

What to coverSuggested preparation: Discuss time-aware validation, appropriate baselines, and how forecast errors affect decisions. Explain how you would detect degradation and decide when a model needs updating.

What trade-offs would you consider in a personalization or recommendation system?

What to coverSuggested preparation: Contrast relevance with coverage and the needs of users or content with limited history. Describe how you would evaluate the system beyond an offline score and check whether it improves the intended outcome.

How would you architect a RAG pipeline over structured and unstructured enterprise data?

What to coverSuggested preparation: Walk through ingestion, embeddings, retrieval, and generation, including how structured records and documents might be used together. Discuss retrieval-quality checks, access controls, and the trade-offs among relevance, latency, and cost.

How would you improve prompts for reliable summarization or metadata enrichment?

What to coverSuggested preparation: Explain how structured outputs, few-shot examples, and systematic prompt comparisons could improve consistency. Describe tests for unsupported claims and what guardrails you would add before using outputs in a workflow.

When would you adapt a foundation model rather than rely on prompting or RAG?

What to coverSuggested preparation: Compare the approaches against the task, available data, evaluation results, operational cost, and maintenance burden. If adaptation is justified, explain how you would assess a parameter-efficient method such as LoRA or adapters.

How would you evaluate an LLM application before and after deployment?

What to coverSuggested preparation: Propose task-specific checks for response quality, hallucinations, and bias alongside latency and cost measures named in the posting. Explain how you would use representative test cases and investigate disagreements between automated scores and human review.

How would privacy and model governance shape an enterprise AI solution?

What to coverSuggested preparation: Discuss data access, sensitive information in prompts or retrieval results, and how to document evaluation and model changes. Explain when you would involve the Responsible AI and Governance teams identified in the posting.

How would you productionize and version an ML or GenAI solution in a cloud environment?

What to coverSuggested preparation: Describe collaboration with Data Engineering and DevOps on deployment, experiment tracking, versioning, and CI/CD. Explain how you would test changes and make an automated retraining decision safely.

How would you detect and respond to model drift, prompt drift, or data-integrity failures?

What to coverSuggested preparation: Distinguish changes in inputs, outputs, and measured performance, then describe useful alerts and diagnostic checks. Explain how reusable monitoring or automation could reduce manual investigation without obscuring failures.

Behavioural practice

Tell me about a time you translated an ambiguous business problem into a measurable data-science solution.

What to coverSuggested preparation: Give the situation, how you aligned Product, Engineering, or business stakeholders on the objective, and the actions you took. State the measured result and what you learned if the outcome differed from expectations.

Describe a time you resolved a disagreement about an ML or AI approach.

What to coverSuggested preparation: Identify the situation and competing priorities, then explain the evidence and trade-offs you brought to the decision. Close with the action taken and its result.

Tell me about a production model or AI-system issue you investigated.

What to coverSuggested preparation: Describe the situation, the signals that exposed the issue, and your actions with relevant partners. Explain the result, including any change to validation, monitoring, or deployment practices.

Describe a time you addressed a privacy, bias, or reliability concern in an AI project.

What to coverSuggested preparation: Set out the concern and your responsibility, then explain how you evaluated it and worked with appropriate stakeholders. Give the resulting decision and what it changed about the solution.

Tell me about a reusable tool or process you built to reduce repeated ML or analytics work.

What to coverSuggested preparation: Explain the recurring problem, the design and adoption actions you took, and how you balanced reuse against team-specific needs. Report the result using evidence such as time saved, reliability, or uptake where available.

Sharpen your knowledge

Topics to prepare

Model selection and validation
The posting calls for statistical and predictive models across forecasting, personalization, and content performance, with rigorous validation. Practice suggestion: compare when gradient boosting, time-series methods, or recommendations fit a proposed use case; specify a baseline, validation approach, and business metric.
RAG architecture and retrieval quality
The role includes RAG over structured and unstructured enterprise data, using embeddings and vector stores. Practice suggestion: diagram a retrieval-to-answer flow and explain how you would test whether poor answers stem from retrieval or generation.
LLM quality and operating trade-offs
The posting explicitly names hallucination, bias, response quality, latency, and cost in production evaluation. Practice suggestion: define measures for each and explain how you would choose between two models when quality improves but latency and cost rise.
Responsible AI and governance
The role works with Responsible AI and Governance teams on privacy, enterprise policies, and ethical practices. Practice suggestion: review a hypothetical knowledge-retrieval use case and identify privacy risks, output guardrails, and questions to resolve with those teams.
Model lifecycle and measurable impact
The posting emphasizes production-grade solutions, experiment tracking, versioning, retraining, drift monitoring, and measurable enterprise value. Practice suggestion: outline how a model would move from experiment to monitored deployment and how its business impact would be assessed.

Skills to revise

Implement a retrieval pipeline
The posting seeks hands-on RAG, embeddings, semantic search, and vector-store experience. Practice suggestion: assemble a small structured-and-unstructured document example, retrieve relevant passages, generate grounded answers, and inspect failed retrievals.
Optimize prompts and structured outputs
Prompt engineering, few-shot learning, structured output design, and systematic optimization are explicit qualifications. Practice suggestion: revise a metadata-enrichment prompt against a fixed set of examples; compare output consistency before and after the changes.
Evaluate LLM outputs
The role calls for evaluation pipelines and guardrails for reliable production outputs. Practice suggestion: create a small test set for summarization or knowledge retrieval, score answer quality and unsupported claims, and record latency and cost alongside the results.
Adapt a foundation model
The posting names parameter-efficient adaptation, including LoRA and adapters, where appropriate. Practice suggestion: sketch or run a small adaptation experiment and compare it with prompt-only and RAG approaches; explain when the added complexity is justified.
Validate a predictive model
The role requires robust statistical and machine-learning models for uses such as forecasting and customer analytics. Practice suggestion: build a baseline and a gradient-boosting model for a sample use case, check for leakage, and compare performance against the stated business objective.
Plan production monitoring
The posting specifies collaboration with Data Engineering and DevOps, CI/CD, versioning, automated retraining, and monitoring for model or prompt drift and data integrity. Practice suggestion: draft a deployment checklist with versioned inputs, monitoring signals, and conditions for investigation or retraining.

Your preparation checklist

Map the role to the posting
The posting describes a senior individual contributor serving Studios, Streaming, Linear Networks, Consumer Products, Games and Direct-to-Consumer. Prepare a one-page map of the business problems and stakeholder groups you would want to understand first.
Build a production ML walkthrough
The role calls for deployed models with measurable business impact. Prepare one project walkthrough covering the problem, model choice, validation, deployment, monitoring, your technical decisions and measured outcome.
Sketch a GenAI system design
The posting calls for RAG across structured and unstructured data. As practice, diagram a hypothetical content-metadata enrichment system: data sources, embeddings and retrieval, prompts and structured outputs, guardrails, and when model adaptation might be warranted.
Draft an evaluation plan
The role evaluates hallucinations, bias, response quality, latency and cost. Prepare a compact test plan with example metrics, failure cases and a method for comparing a baseline with a proposed improvement.
Explain the production lifecycle
The posting specifies cloud productionization, CI/CD, experiment tracking, versioning, retraining and drift monitoring. Prepare an architecture sketch showing handoffs with Data Engineering and DevOps, data-integrity checks and a reusable component.
Organize evidence for the stated qualifications
The posting specifies a quantitative degree, 8+ years of relevant data science experience, 2+ years of GenAI experience and production AI/ML impact. Prepare a dated project timeline and concise evidence for each requirement, without stretching what the evidence shows.
Write questions for the employer
Prepare questions about which portfolio use cases are priorities, how Product and Engineering define measurable impact, and how Responsible AI and Governance teams review privacy and model risks. These are questions to ask, not assumptions about WBD’s processes.

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…