New opportunity

Senior Data Scientist

General Mills · Mumbai, India

About this role

General Mills’ Digital and Technology India team seeks a Senior Data Scientist to lead scalable data science solutions for high-value business problems in the consumer packaged goods (CPG) domain. The role spans core machine learning, deep learning, generative AI and agentic AI, including enterprise intelligent automation. It is a hybrid position in Powai, Mumbai, Maharashtra.

Responsibilities

Partner with business stakeholders to define problems and measurable outcomes, communicate insights, manage delivery commitments and advise teams across functions.
Develop and evaluate regression, classification, forecasting, optimization, NLP and deep-learning models using experimentation and error analysis. Design agentic AI, LLM and RAG solutions covering data curation, ingestion, chunking, embeddings, vector indexing, retrieval optimization, fine-tuning and human-in-the-loop evaluation.
Maintain ontologies, taxonomies and metadata for consistent enterprise retrieval and reasoning. Work with engineering teams to deploy, monitor and improve production models using MLOps practices; address performance degradation and maintainability. Apply privacy, security, governance, transparency and bias-reduction practices.
Provide technical reviews and guidance, establish data science practices, mentor junior staff, interns and contractors, evaluate tools and pipelines, and pursue business-aligned applied research and relevant open-source innovation.

Minimum qualifications:

10–12 years of relevant analytics or data science experience, including delivery of production ML and AI solutions. Bachelor’s or master’s degree from a Tier 1 institute in computer science, artificial intelligence, machine learning, statistics, applied mathematics, operations research, industrial engineering or a related quantitative field.
Strong supervised-learning expertise, exposure to unsupervised learning and NLP, and knowledge of MLOps, containerization, data lineage, monitoring and visualization. Proficiency in Python, SQL and Google Cloud Platform; experience with Agile delivery and strong stakeholder management, consulting, communication and storytelling skills. Experience in FMCG/CPG is listed among the minimum qualifications; supply chain analytics is preferred.

Preferred qualifications

Leadership of end-to-end generative AI, LLM, NLP, agentic AI or RAG implementations; expertise in RAG workflows, hypothesis-driven experimentation and human-in-the-loop evaluation. Experience with tools such as Vertex AI, LangChain, LangGraph, Google ADK or vector databases; deep learning, optimization or operations research; GCP architecture, enterprise-scale AI design, cybersecurity fundamentals and Responsible AI governance.

Applicants must meet the minimum age requirements in the country where the job is located.

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

Skills for this role

Data ScienceMachine LearningGenerative AIAgentic AILarge Language ModelsRetrieval-Augmented GenerationNatural Language ProcessingDeep LearningRegressionClassificationDecision TreesEnsemble MethodsTime Series ForecastingNeural NetworksOptimizationUnsupervised LearningModel EvaluationExperimentationError AnalysisData CurationData IngestionChunkingEmbeddingsVector IndexingRetrieval OptimizationFine-TuningHuman-in-the-Loop EvaluationOntologiesTaxonomiesMetadata Management��

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

Frame CPG problems as measurable outcomes
The role partners with business stakeholders to define high-value problems, communicate recommendations, and manage delivery commitments across functions.
Lead rigorous model development
The posting calls for scalable solutions across forecasting, classification, optimization, NLP, and deep learning, supported by experimentation, evaluation, and error analysis.
Design end-to-end GenAI and retrieval solutions
Responsibilities span agentic AI, LLMs, and RAG—from data curation and retrieval optimization to human-in-the-loop evaluation—and maintaining ontologies, taxonomies, and metadata for consistent enterprise reasoning.
Own production quality and responsible AI
The role works with engineers to deploy, monitor, and improve models, address degradation and maintainability, and apply privacy, security, governance, transparency, and bias-reduction practices.
Set technical direction through others
The senior remit includes solution reviews, constructive technical challenge, mentoring, and shaping data science practices as the team evaluates tools and pipelines.

Likely challenges to think through

Balance technical ambition with measurable value
A likely preparation challenge is explaining when a simpler ML approach is preferable to GenAI or agentic AI for a CPG problem. Prepare a decision framework covering outcome metrics, evaluation evidence, complexity, and delivery commitments; this is an inferred trade-off, not a stated employer issue.
Make enterprise retrieval both consistent and useful
RAG quality may require balancing retrieval relevance against consistent business semantics and human review. Prepare to explain how chunking, indexing, metadata, error analysis, and reviewer feedback would inform changes; the posting specifies these responsibilities, not a current retrieval problem.
Sustain quality after deployment
Production performance can change while privacy, security, and maintainability still matter. Prepare an example of monitoring, diagnosing degradation, coordinating with engineers, and deciding when to retrain or revise a solution; this is a plausible role challenge, not a confirmed incident.

Practice your answers

Technical practice

How would you turn a CPG stakeholder's business problem into a measurable data science objective?

What to coverThe role calls for problem framing and measurable outcomes. Suggested answer focus: define the decision the solution would support, agree on a business outcome and evaluation measure, and explain how you would communicate results and limitations.

How would you choose and evaluate a model for a regression or classification problem?

What to coverThe posting names regression, decision trees, ensembles, neural networks and model tuning. Suggested answer focus: compare a simple baseline with more complex approaches, justify metrics and validation, and use error analysis to decide whether added complexity helps.

How would you design a forecasting evaluation that reflects future performance?

What to coverTime-series forecasting is explicitly listed. Suggested answer focus: explain time-ordered validation, a relevant baseline, leakage checks and how you would examine errors across forecast horizons or important segments.

When would you use optimization rather than prediction for a business decision?

What to coverOptimization is a stated responsibility, and supply chain analytics is preferred. Suggested answer focus: distinguish predicting an outcome from selecting an action, then discuss objectives, constraints, uncertain inputs and how you would assess a proposed decision.

How would you structure an experiment and diagnose model failures?

What to coverThe posting emphasizes hypothesis-driven experimentation, evaluation strategies and error taxonomies. Suggested answer focus: state a testable hypothesis, select comparison measures, categorize failures and explain how findings would guide the next iteration.

How would you design and evaluate an enterprise RAG pipeline?

What to coverThe role explicitly covers ingestion, chunking, embeddings, vector indexing, retrieval optimization and human-in-the-loop evaluation. Suggested answer focus: explain how you would test retrieval and response quality separately, then weigh changes to chunking, retrieval and fine-tuning against cost and maintainability.

What would justify an agentic AI workflow instead of a simpler LLM solution?

What to coverThe role includes agentic AI and enterprise intelligent automation. Suggested answer focus: compare the control and complexity of the two approaches, and describe how you would limit actions, assess failures and incorporate human review.

How do ontologies, taxonomies and metadata improve enterprise retrieval?

What to coverMaintaining business semantics for consistent retrieval and reasoning is a stated duty. Suggested answer focus: explain how consistent terms and metadata help identify relevant information, and how you would detect and resolve conflicting definitions.

How would you deploy and maintain a model when production performance degrades?

What to coverThe posting names collaboration with engineers, monitoring, data lineage, containerization and GCP. Suggested answer focus: discuss reproducible deployment, monitoring data and model behaviour, diagnosing degradation, and deciding when to revise or roll back a model.

How would you apply Responsible AI controls to an ML or generative AI solution?

What to coverThe role explicitly calls for privacy, security, governance, transparency and bias reduction. Suggested answer focus: identify risks during data preparation, evaluation and deployment, then describe proportionate controls and evidence that they work.

Behavioural practice

Tell me about a time you aligned stakeholders on a measurable outcome for a complex analytics problem.

What to coverSuggested answer focus: use situation, action and result to show how you clarified competing needs, agreed on a useful measure and communicated the recommendation. Keep the example grounded in what you actually did.

Tell me about a time you constructively challenged another technical approach.

What to coverThe role includes solution reviews and consultation. Suggested answer focus: describe the situation, the evidence and trade-offs you raised, your actions with the team, and the resulting decision.

Tell me about a time you helped a junior colleague develop technically or in a business domain.

What to coverMentoring junior data scientists, interns and contractors is part of the role. Suggested answer focus: describe the situation, the guidance you provided without taking over, and an observable result.

Tell me about a time priorities changed while you had delivery commitments.

What to coverThe posting calls for priority management and Agile delivery. Suggested answer focus: explain the situation, how you re-estimated or negotiated scope and communicated changes, and what resulted.

Tell me about a time you evaluated a new method or tool against a business need.

What to coverThe role includes evaluating platforms and pipelines and pursuing business-aligned applied research. Suggested answer focus: describe the need, the criteria and actions you used to assess alternatives, and the resulting decision or learning.

Sharpen your knowledge

Topics to prepare

Supervised learning and forecasting
The posting requires strong expertise in regression, decision trees, ensembles, time-series forecasting, neural networks and model tuning. Revision exercise: compare two approaches to a hypothetical CPG forecasting problem, explaining validation choices, likely errors and when a more complex model is justified.
Experimentation and model evaluation
The role calls for robust experimentation, evaluation and error analysis. Revision exercise: write a test plan for a classification model that defines a measurable outcome, baseline, evaluation measures and an error taxonomy.
RAG, agentic AI and business semantics
The accountabilities cover ingestion, chunking, embeddings, vector indexing, retrieval optimization and human-in-the-loop evaluation, alongside ontologies, taxonomies and metadata. Revision exercise: diagram an enterprise retrieval workflow and explain how its metadata and review process support consistent answers.
Production ML and Responsible AI
MLOps, containerization, data lineage, monitoring and visualization are minimum-qualification areas; the role also applies privacy, security, transparency and bias-reduction practices. Revision exercise: review a proposed model deployment and identify what to monitor, how to investigate degradation and which governance checks belong in delivery.
CPG problem framing and technical leadership
CPG experience, measurable stakeholder outcomes, solution reviews and mentoring are stated expectations; supply chain analytics is preferred. Revision exercise: frame a hypothetical CPG decision as an analytics problem, define success measures and prepare two constructive questions for a design review. The hypothetical problem is preparation, not an advertised project.

Skills to revise

SQL data preparation
SQL proficiency is a minimum qualification. Practice writing queries that join and aggregate hypothetical CPG data into a modeling dataset, then check missing values, duplicate keys and potential data leakage.
Python model evaluation
Python proficiency and delivery of production ML solutions are required. Practice training and comparing a simple baseline and a supervised model, then produce a compact error analysis and explain the tuning decision.
RAG workflow prototyping
Designing RAG solutions is an accountability; hands-on experience with Vertex AI, LangChain, LangGraph, Google ADK or vector databases is preferred. As a preparation exercise, prototype ingestion, chunking and retrieval on sample documents, then inspect retrieval failures and record human-review findings. No particular framework is mandatory in the posting.
GCP deployment walkthrough
GCP proficiency and production ML practices are minimum qualifications. Practice walking through how a model would be packaged, deployed and monitored on GCP, documenting data lineage, a degradation signal and a maintenance response; do not assume a specific employer architecture.
Stakeholder communication
The posting requires stakeholder management, storytelling and communication with technical and non-technical audiences. Practice a brief explanation of a hypothetical model recommendation that states the business measure, evidence, uncertainty and delivery trade-off, then adapt it for a technical solution review.

Your preparation checklist

Map evidence to the stated qualifications
Prepare a one-page inventory of your actual production ML and AI deliveries, FMCG/CPG work, Python, SQL, GCP, Agile delivery, and relevant education. Mark supply chain analytics and GenAI tooling separately as preferred areas; do not imply experience you cannot substantiate.
Build a business-to-production project walkthrough
Choose a real CPG or other relevant project and prepare a five-minute account of stakeholder problem framing, measurable outcomes, your decisions, delivery commitments, production results, and what you would change. The posting emphasizes senior ownership and communication across functions.
Rehearse a core ML evaluation example
Prepare one concrete example covering model choice, baselines, experimentation, performance tuning, error analysis, and communication of results. Be ready to explain trade-offs for forecasting, regression, or classification rather than merely listing algorithms.
Sketch an enterprise RAG or agentic AI design
As an interview exercise, draw a solution from data curation and ingestion through chunking, embeddings, indexing, retrieval, evaluation, and deployment. Explain where ontologies, taxonomies, metadata, human review, and any justified fine-tuning fit; these are responsibilities in the posting, not a stated interview assignment.
Document production and Responsible AI decisions
Create a short architecture and operations note showing how you would use GCP, lineage, containerization, monitoring, and responses to performance degradation. Include privacy, security, governance, transparency, and bias checks relevant to the proposed solution.
Prepare leadership and delivery stories
Select specific examples of solution reviews, constructive technical challenge, mentoring, stakeholder influence, and Agile estimation or priority trade-offs. For each, record your actions, the decision reached, and an observable outcome.
Draft questions for the employer
Bring questions about which CPG problems have priority, how success is measured, how data science works with ML and systems engineers, and what standards govern enterprise retrieval and Responsible AI. These ask about the advertised work without assuming a particular internal process.

Availability last checked: 5 Nov 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 India. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…