New opportunity

Manager-Data Science

Cognizant · Hyderabad, India

About this role

Cognizant seeks a Manager-Data Science to serve as a Data Science & AI/ML Lead on its AI Engineering team. The role leads the design, development and optimization of enterprise AI solutions, turning experimental models into scalable, production-grade capabilities. The position is hybrid in Hyderabad, India; the posting does not specify the number of office days.

Responsibilities

Lead AI and machine learning development, from data preparation and model training through fine-tuning and deployment strategy.
Design scalable pipelines to curate, cleanse and prepare structured and unstructured data for machine learning and foundation model training.
Build agentic AI workflows for automation and decision-making. Architect retrieval-augmented generation (RAG) solutions using semantic search, vector embeddings, re-ranking and optimized document retrieval.
Establish evaluation and benchmarking frameworks for model accuracy, reliability, safety and performance. Collaborate with architects, data engineers, cloud specialists and business stakeholders on cloud-native AI deployments.

Required qualifications

10+ years of experience in Data Science, Machine Learning or AI Engineering, including preparation and management of large-scale datasets.
Expert-level Python proficiency and experience with machine learning frameworks such as PyTorch, TensorFlow and Scikit-learn.
Hands-on experience with large and small language models, prompt engineering and agentic AI frameworks; knowledge of vector databases, semantic search architectures, and SQL and NoSQL data platforms.
Experience deploying and operationalizing models on cloud AI platforms such as Azure AI Foundry, Vertex AI or Amazon SageMaker.

Preferred qualifications

Advanced RAG architectures and enterprise search; LoRA, PEFT and hyperparameter optimization for model fine-tuning; and evaluation tools such as RAGAS or TruLens.
Synthetic data generation, data labeling, responsible AI and bias mitigation, plus deployment in cloud-native, containerized environments on Azure, AWS or Google Cloud Platform.

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

Skills for this role

Data ScienceMachine LearningAI EngineeringPythonPyTorchTensorFlowScikit-learnData PipelinesData PreparationModel TrainingModel Fine-tuningModel DeploymentLarge Language ModelsSmall Language ModelsPrompt EngineeringAgentic AIRetrieval-Augmented GenerationSemantic SearchVector EmbeddingsRe-rankingVector DatabasesSQLNoSQLAzure AI FoundryVertex AIAmazon SageMakerModel EvaluationBenchmarkingLoRAPEFT; Hyperparameter Optimization

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

Lead AI from development to deployment
The posting calls for leading data preparation, model training, fine-tuning and deployment strategy, with the aim of turning experimental models into production-grade capabilities.
Build scalable data pipelines
Design and implement pipelines that curate, cleanse and prepare large-scale structured and unstructured data for machine learning and foundation model training.
Develop agentic workflows
Build and optimize agentic AI workflows using orchestration frameworks for automation and decision-making.
Architect retrieval-augmented generation
Design RAG solutions spanning semantic search, vector embeddings, re-ranking and document retrieval optimization.
Evaluate and operationalize AI
Establish benchmarks for accuracy, reliability, safety and performance while working with architects, data engineers, cloud specialists and business stakeholders on cloud-native deployments.

Likely challenges to think through

Moving beyond a successful experiment
Likely preparation challenge, not a stated employer problem: a promising model may be difficult to scale and operate reliably. Prepare an example of how you would choose a deployment strategy and coordinate technical and business stakeholders without losing sight of model quality.
Balancing retrieval quality against complexity
Likely preparation challenge: embeddings, re-ranking and retrieval optimization can improve relevance while adding latency and operational complexity. Prepare to explain how you would compare retrieval designs using measurable quality and performance criteria.
Testing agentic behavior safely
Likely preparation challenge: greater workflow autonomy can make reliability and safety harder to assess. Prepare an evaluation approach that tests decision quality, failure cases and safety alongside performance, rather than relying on a single accuracy measure.

Practice your answers

Technical practice

How would you design a scalable data pipeline for structured and unstructured AI training data?

What to coverThe role includes curating, cleansing and preparing large-scale datasets for machine learning and foundation models. Suggested answer focus: explain data-quality checks, processing at scale and how you would make prepared data usable for training.

How would you take a machine-learning model from training to a deployment decision?

What to coverThe posting calls for leadership across model training, optimization and deployment strategy, with expert Python and experience in frameworks such as PyTorch, TensorFlow and Scikit-learn. Suggested answer focus: explain your choice of framework, validation approach and criteria for moving beyond an experiment.

When would you choose a small language model rather than a large one?

What to coverHands-on work with both LLMs and SLMs, including prompt engineering, is listed. Suggested answer focus: compare task quality, resource demands and prompt-based approaches before recommending a model.

How would you design an agentic workflow for automation and decision-making?

What to coverBuilding and optimizing agentic AI workflows with orchestration frameworks is a stated duty. Suggested answer focus: describe how tasks and decisions are coordinated, how failures are handled and when a simpler workflow would be preferable.

How would you improve retrieval quality in a RAG system?

What to coverThe role specifically covers semantic search, embeddings, re-ranking and document-retrieval optimization. Suggested answer focus: explain how you would diagnose poor retrieval, compare improvements at each stage and test whether answers improve.

How would you choose data stores for an enterprise search architecture?

What to coverThe posting requests knowledge of vector databases, semantic search architectures and SQL/NoSQL platforms. Suggested answer focus: explain what each store would hold and weigh retrieval needs against data management and operational complexity.

How would you decide between prompting and model fine-tuning?

What to coverModel fine-tuning is part of the role; LoRA, PEFT and hyperparameter optimization are listed as preferred expertise. Suggested answer focus: compare expected quality gains with data, compute and maintenance costs, and describe how you would test the decision.

How would you benchmark an AI solution before production deployment?

What to coverThe role calls for evaluation frameworks covering accuracy, reliability, safety and performance; RAGAS and TruLens are named as examples of preferred tools. Suggested answer focus: define task-relevant measures, representative test data and how results would guide release decisions.

How would you assess synthetic or labeled data for responsible AI use?

What to coverSynthetic data generation, labeling strategies and bias mitigation appear among the preferred qualifications. Suggested answer focus: explain how you would check data quality and representation, then evaluate whether the resulting model behaves reliably across relevant groups.

What would you consider when operationalizing a model on a cloud AI platform?

What to coverThe posting names Azure AI Foundry, Vertex AI and Amazon SageMaker, and mentions cloud-native, containerized deployment. Suggested answer focus: discuss deployment and monitoring considerations, scaling trade-offs and coordination with cloud and data engineering specialists without assuming a particular platform is mandated.

Behavioural practice

Tell me about a time you led an AI project from experiment toward production.

What to coverThe role emphasizes turning experimental models into scalable capabilities. Suggested answer focus: use situation, action and result to show how you set priorities, addressed a production-readiness obstacle and measured the outcome.

Describe a time you coordinated specialists to deliver an AI solution.

What to coverThe role involves architects, data engineers, cloud specialists and business stakeholders. Suggested answer focus: use situation, action and result to explain how you aligned responsibilities, resolved a dependency and kept delivery moving.

Tell me about a time you had to explain an AI design trade-off to a business stakeholder.

What to coverThe posting combines technical design with collaboration on AI-powered solutions. Suggested answer focus: use situation, action and result to show how you presented options, incorporated stakeholder needs and reached a decision.

Describe a time evaluation revealed a reliability or safety concern.

What to coverEstablishing measures for model reliability and safety is a stated responsibility. Suggested answer focus: use situation, action and result to explain what the evidence showed, how you responded and what changed afterward.

Sharpen your knowledge

Topics to prepare

From experimental model to production capability
The posting calls for leading model training, fine-tuning and deployment strategy. Suggested exercise: sketch the decisions and handoffs needed to take a model from an experiment to a scalable service; compare full fine-tuning with LoRA or PEFT, which the posting lists as preferred.
Large-scale data preparation
The role includes managing large datasets and designing pipelines for structured and unstructured data. Suggested exercise: explain how you would curate, cleanse and validate both data types before model training, including where SQL and NoSQL platforms fit.
LLM, SLM and agentic workflow design
Hands-on experience with large and small language models, prompt engineering and agentic frameworks is required. Suggested exercise: compare when to use an LLM or SLM, then diagram an agentic workflow for a hypothetical decision task, including failure handling.
RAG and enterprise search architecture
The role explicitly covers semantic search, embeddings, re-ranking and document retrieval optimization; advanced RAG is preferred. Suggested exercise: trace a query through indexing, retrieval, re-ranking and answer generation, explaining the trade-offs at each stage.
Evaluation, reliability and responsible AI
The role establishes benchmarks for accuracy, reliability, safety and performance; RAGAS or TruLens and bias mitigation are preferred. Suggested exercise: define a small evaluation set and explain which measures and failure cases you would review before deployment.
Cloud AI operationalization
Experience deploying and operationalizing models on Azure AI Foundry, Vertex AI or Amazon SageMaker is required. Suggested exercise: outline a deployment and monitoring plan on one listed platform, identifying decisions to coordinate with architects, data engineers and cloud specialists.

Skills to revise

Python model development
Expert Python proficiency and experience with frameworks such as PyTorch, TensorFlow and Scikit-learn are requested. Suggested exercise: implement a compact training and evaluation workflow in Python using a listed framework, and explain how you would make its results reproducible.
Building a preparation pipeline
The posting calls for scalable curation and cleansing of structured and unstructured data. Suggested exercise: write a small Python pipeline that ingests both types, checks data quality and produces training-ready outputs; explain how its design would change at scale.
Implementing retrieval and re-ranking
RAG implementation and knowledge of vector databases are central to the role. Suggested exercise: build a small document-retrieval flow using embeddings and semantic search, add re-ranking, and compare the retrieved results before and after.
Constructing an agentic workflow
The role builds and optimizes agentic AI workflows using orchestration frameworks. Suggested exercise: prototype a short multi-step workflow with a clear decision point, then test what happens when a step returns an unreliable result.
Running a model and RAG benchmark
The role establishes evaluation frameworks; RAGAS and TruLens are listed as preferred tools. Suggested exercise: run the same test set across two prompt or retrieval variants, record accuracy, reliability, safety and performance findings, and justify which variant you would advance.
Planning a cloud deployment
Model deployment and operationalization on a listed cloud AI platform are required. Suggested exercise: prepare a platform-specific deployment walkthrough covering model artifacts, release checks and operational monitoring; use a design walkthrough if platform access is unavailable.

Your preparation checklist

Map your evidence to the role
Prepare a one-page evidence map for the posting’s 10+ years of experience, large-scale datasets, Python, ML frameworks, language models and cloud AI deployment. Use specific examples where available; mark any gaps rather than inventing experience.
Rehearse an end-to-end project walkthrough
Choose a relevant project, if you have one, and explain your leadership decisions from data preparation and model training through fine-tuning, deployment and operationalization. Bring a simple architecture diagram and measurable outcomes you can substantiate.
Sketch a scalable data pipeline
As a preparation exercise, diagram how you would curate, cleanse and prepare structured and unstructured data for ML or foundation-model training. Note data-quality checks and where SQL or NoSQL platforms would fit.
Design a RAG and agentic workflow
Prepare a whiteboard design showing document retrieval, embeddings, semantic search and re-ranking, plus an agent workflow for automation or decision-making. Explain your design trade-offs rather than presenting it as a Cognizant project.
Build an evaluation plan
Draft a small benchmark plan covering accuracy, reliability, safety and performance. Include sample test cases, failure analysis and, if relevant to your experience, how RAGAS or TruLens could be used.
Explain production and fine-tuning choices
Prepare a comparison of model and prompt choices, optional LoRA or PEFT fine-tuning, and deployment on one of the cloud AI platforms named in the posting. Be ready to discuss scaling and operational trade-offs without claiming experience you do not have.
Write questions for the hiring team
Ask how this AI Engineering lead collaborates with architects, data engineers, cloud specialists and business stakeholders; what production and evaluation priorities matter most; and what the Hyderabad hybrid arrangement entails, since the posting leaves office-day frequency unspecified.

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…