New opportunity

Data Scientist II, International Machine Learning

Amazon · Gurugram, India

About this role

Amazon’s International Machine Learning team is seeking a Data Scientist II in Gurugram to develop scalable machine learning solutions for its international consumer businesses. The team focuses on problems in India and other emerging markets across MENA and Latin America. Its work uses large-scale business data to improve customer experience, automate processes, reduce costs and support business growth.

Responsibilities

Apply machine learning and analytical techniques to business problems; extract insights from historical data to automate and optimize key processes.
Design, develop, evaluate and deploy scalable ML models, working with software engineers on real-time production implementations.
Build efficient, automated processes for large-scale analysis, model development, validation and maintenance.
Partner with business stakeholders, product managers and engineers to identify opportunities, propose solutions and implement complex models. Lead projects and mentor scientists and engineers in ML techniques.

Basic qualifications: At least 2 years of data scientist experience; 3+ years of experience with data-querying languages such as SQL, scripting languages such as Python, or statistical and mathematical software such as R, SAS or Matlab; and 3+ years of experience with machine learning or statistical-modeling tools, techniques and performance factors. Candidates need experience applying theoretical models in practice, using quantitative analysis for business decisions, developing ML models across the credit lifecycle—including pre-underwriting, underwriting, post-underwriting and collections—and communicating complex concepts effectively. Fintech, lending or credit-risk experience is highly preferred.

Preferred qualifications

A Ph.D. in a STEM field; knowledge of ML concepts for reasoning and problem-solving; experience in an ML or data scientist role at a large technology company; and the ability to turn insights into strategy and communicate them clearly in writing. Three or more years of credit-lifecycle ML model development and 3+ years in fintech, lending or credit risk are also preferred.

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

Skills for this role

Machine LearningStatistical ModelingData AnalysisSQLPythonRSASMATLABQuantitative AnalysisCredit RiskModel DevelopmentModel ValidationModel DeploymentData-Driven Decision MakingWritten CommunicationVerbal CommunicationMentoring

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

Tie ML work to business outcomes
The posting calls for owning business problems and metrics, using quantitative analysis to guide decisions, and improving customer experience, automation, cost or growth. Prepare to explain how a model's results would translate into a business decision.
Deliver models into production
The role includes designing, evaluating and deploying scalable models with software engineers, including real-time implementations. Focus preparation on decisions from model design through deployment, not just offline performance.
Build repeatable analysis and model processes
The team seeks automated processes for large-scale data analysis, model development, validation and maintenance using historical business data. Be ready to discuss how those stages remain efficient and reliable at scale.
Apply ML across the credit lifecycle
The basic qualifications specify model development across pre-underwriting, underwriting, post-underwriting and collections; fintech, lending or credit-risk experience is highly preferred. Prepare to distinguish the modeling decisions and evaluation needs at different stages.
Lead through cross-functional communication
The role partners with business stakeholders, product managers and engineers, leads projects, and mentors scientists and engineers. Prepare concise explanations of model choices, business implications and implementation trade-offs for different audiences.

Likely challenges to think through

Balancing local relevance with reusable solutions
Preparation hypothesis: An India-first team also addressing markets across MENA and Latin America may need to balance market-specific patterns against scalable approaches. Prepare to explain how you would test whether an approach transfers before reusing it.
Reconciling model quality with real-time operation
Preparation hypothesis: A model that performs well in analysis may be difficult to run or maintain in a real-time system. Prepare to discuss trade-offs among predictive performance, implementation complexity and operational reliability with engineers.
Evaluating credit models at different lifecycle stages
Preparation hypothesis: Pre-underwriting, underwriting and collections can have different decisions and outcome timelines, making one evaluation approach insufficient. Prepare to explain how you would choose stage-appropriate measures and validate models before use.

Practice your answers

Technical practice

How would you decide whether a business problem needs an ML solution?

What to coverThe role calls for identifying business problems and using ML and analysis to improve customer experience, cost, automation or growth. Preparation suggestion (inferred): explain how you would define the decision and success metric, then compare an ML approach with a simpler alternative.

How would you turn large volumes of historical data into a reliable modeling dataset?

What to coverThe posting emphasizes extracting information from large amounts of historical business data and names querying or scripting experience. Preparation suggestion (inferred): discuss joins, data quality checks and how you would prevent information unavailable at decision time from entering the dataset.

How would you choose and tune a statistical or ML model for an applied problem?

What to coverThe role requires applying theoretical models in practice and understanding factors that affect model performance. Preparation suggestion (inferred): compare a simple baseline with a more complex model, covering fit, interpretability, computational cost and validation.

How might modeling objectives change across the credit lifecycle?

What to coverThe qualifications explicitly mention model development for pre-underwriting, underwriting, post-underwriting and collections. Preparation suggestion (inferred): distinguish the decision, available data and outcome at each stage rather than treating the lifecycle as one prediction task.

How would you evaluate a model intended to safeguard financial transactions?

What to coverThe page gives transaction risk models as an example of the team's practical ML problems. Preparation suggestion (inferred): explain the trade-off between missed risky transactions and unnecessary intervention, and how you would connect model measures to the business decision.

How would you establish whether a model improves a business metric?

What to coverThe role includes evaluating models, making data-driven decisions and owning business problems and metrics. Preparation suggestion (inferred): separate predictive performance from business impact, and explain what comparison and validation would support a decision to deploy.

What would you work through with engineers before a real-time model deployment?

What to coverThe responsibilities include working with software engineers on real-time implementations of scalable models. Preparation suggestion (inferred): cover when inputs become available, response-time constraints, prediction consistency and how the implementation would handle failures.

How would you automate model validation and maintenance at scale?

What to coverThe posting calls for efficient, automated processes for large-scale analysis, development, validation and maintenance. Preparation suggestion (inferred): describe repeatable checks, monitoring for changing data or performance, and criteria for investigating or updating a model.

How would you approach extracting product attributes from mixed data?

What to coverThe page cites catalog-quality improvements through attribute extraction from structured and unstructured data as an example problem. Preparation suggestion (inferred): discuss how you would define the attribute, obtain evaluation data and handle ambiguous or missing information at scale.

How would you evaluate a product-bundle recommendation approach?

What to coverThe page names bundle recommendations across millions of products as another example application. Preparation suggestion (inferred): explain how you would assess recommendation relevance alongside customer and business outcomes, including the limits of relying only on historical interactions.

Behavioural practice

Tell me about a time you identified a worthwhile ML opportunity with business partners.

What to coverThe role involves partnering with business stakeholders to identify problems and propose solutions. Use a real situation, your actions in defining the problem and the result; preparation suggestion (inferred): explain how you established that the proposed approach served a business need.

Describe a time you worked with engineers to put a model into use.

What to coverThe posting emphasizes close work with engineering teams on production implementations. Give the situation, your specific coordination and the result; preparation suggestion (inferred): address how you resolved a practical difference between modeling goals and implementation constraints.

Tell me about a project you led while helping others apply ML techniques.

What to coverLeading projects and mentoring scientists or engineers are stated responsibilities. Describe the situation, what you personally did to guide the work and the result; preparation suggestion (inferred): distinguish your contribution from the team's.

Describe a time you communicated a complex analysis to support a decision.

What to coverThe qualifications emphasize written and verbal communication and turning insights into strategy. Use a real situation, the communication choices you made and the resulting decision or next step; preparation suggestion (inferred): show how you conveyed uncertainty as well as the main finding.

Tell me about a time you owned a business metric through a changing project.

What to coverThe page highlights end-to-end ownership of business problems and metrics. Set out the situation, how you adapted your actions when evidence or constraints changed, and the result; preparation suggestion (inferred): make clear how you measured progress rather than relying on model performance alone.

Sharpen your knowledge

Topics to prepare

Credit-lifecycle modeling
Posting: Experience developing ML models across pre-underwriting, underwriting, post-underwriting and collections is listed in the basic qualifications. Practice: For a hypothetical lending case, define the decision, prediction target and available data at each stage; explain how the modeling problem changes across stages.
Model performance and validation
Posting: The role calls for ML and statistical-modeling techniques, understanding factors that affect performance, and model evaluation and validation. Practice: Compare two candidate approaches for a hypothetical risk model and explain how you would detect misleading results or performance changes over time.
Scalable, real-time ML systems
Posting: The scientist designs and deploys scalable models, works with engineers on real-time implementations, and establishes automated development and maintenance processes. Practice: Sketch the path from historical data to a production prediction, identifying what must be validated, automated and monitored.
Business decisions across international use cases
Posting: The team tackles India-first and other emerging-market problems; examples include transaction risk, product bundles, catalog attributes and address quality. Practice: Choose one listed use case and define its business decision, success measure and likely trade-off. Treat the proposed measure as interview preparation, not a stated employer metric.

Skills to revise

Query and analyze historical data
Posting: Data querying or scripting experience is listed, and the role extracts information from large amounts of historical business data. Practice: Using a small hypothetical dataset, write a SQL query to assemble a modeling table and check for missing values and duplicate records. SQL is a listed example, not an exclusive requirement.
Build and test an applied model
Posting: The role involves developing ML models and applying theoretical models in practice. Practice: In Python or another listed language or software environment, train a simple model on sample data, evaluate it on held-out data and explain which performance factors you would investigate.
Design repeatable model checks
Posting: Responsibilities include automated large-scale analysis, model validation and maintenance. Practice: Draft a repeatable checklist for input checks, evaluation before deployment and post-deployment performance review; explain which steps you would automate.
Explain and lead a model proposal
Posting: The scientist partners with business and engineering teams, leads projects, mentors others, and communicates complex concepts in writing and verbally. Practice: Prepare a brief written proposal for one listed use case, then explain its decision, evidence, limitations and implementation needs aloud to a nontechnical stakeholder.

Your preparation checklist

Prepare an end-to-end ML project walkthrough
The role covers model design, evaluation and deployment. Prepare one walkthrough showing the business problem, data, modeling choices, validation, production handoff and measured outcome; use a hypothetical case if you cannot discuss past work.
Map a model across the credit lifecycle
The posting explicitly asks for experience spanning pre-underwriting, underwriting, post-underwriting and collections. If you have relevant work, prepare examples for the stages you have covered; otherwise, outline a clearly labeled hypothetical design and its limitations.
Practice a large-scale analysis exercise
The role uses historical business data to automate processes. Prepare a short SQL or Python exercise that defines a business metric, extracts useful signals and explains how you would check data quality and turn the result into a decision.
Sketch a production model design
The posting calls for scalable models, real-time implementations with engineers, and automated validation and maintenance. Draft a design covering inference, evaluation, monitoring and what you would do when model performance changes.
Prepare a leadership and communication example
The role involves partnering with business, product and engineering teams, leading projects and mentoring others. Prepare a concise example of a technical decision, how you explained its trade-offs and how collaborators used the result.
Write targeted questions for the team
Ask which India-first or other emerging-market problems are priorities, how the team measures customer and business impact, and how scientists and engineers share responsibility for models in production.

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…