New opportunity

Gen AI/ ML Data Scientist

Quadratyx · Mumbai, India

About this role

Quadratyx seeks a Gen AI/ ML Data Scientist in Mumbai to own end-to-end delivery of client-facing ML and data capabilities. The role turns business needs into production-grade data pipelines, deployed models and measurable outcomes, with accountability for technical delivery, operational reliability and stakeholder expectations. It reports to a Project Manager.

Responsibilities

Deliver production ML systems for recommendations, NLP and personalization, from problem definition and metric design through deployment, monitoring and improvement based on real-world results.
Architect, build and operate scalable ETL/ELT pipelines that produce reliable analytics- and ML-ready datasets from disparate sources. Establish data architecture standards for data lakes, including partitioning, schema evolution, file optimization and columnar formats such as Parquet.
Build CI/CD-aligned ML workflows covering automated testing, reproducibility, deployment infrastructure and production monitoring. Expose model and data capabilities through versioned APIs and services aligned with microservices architecture.
Lead client problem discovery, scope and constraint definition, technical planning, risk management and communication of measurable success criteria; manage changes and stakeholder expectations.

Requirements

The posting lists 5 years of experience in its role information and specifies 4–6 years of hands-on production data science and data engineering delivery experience in its qualifications. A bachelor’s degree in computer science, Data Analytics, Engineering or a related discipline is required.
Demonstrated ownership of at least one production ML deliverable, such as a recommender, NLP, personalization, segmentation or GenAI feature, including evaluation and post-launch changes. Production experience developing and deploying computer vision or image recognition solutions, such as classification, detection, OCR, segmentation or multimodal vision, using frameworks such as OpenCV, PyTorch, TensorFlow, YOLO or Hugging Face Vision models.
Hands-on delivery of scalable pipelines, with knowledge of volumes, SLAs, latency or freshness, backfills and failure handling; practical data-lake experience, including Parquet or ORC and performance or cost trade-offs; and production software practices covering testing, CI/CD, API design, monitoring and incident learnings. Experience translating client requirements into technical plans and handling scope changes is required.
Strong independent and collaborative working ability, communication with technical and non-technical stakeholders, critical thinking, structured problem-solving, adaptability and ownership. Candidates should communicate data insights as actionable business recommendations in a fast-paced environment.

Interested candidates should email their profiles to jobs@quadratyx.com. Quadratyx states that candidates may be placed or moved to company, partner or customer locations globally.

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

Skills for this role

Machine LearningData ScienceData EngineeringGenerative AINatural Language ProcessingRecommendation SystemsPersonalizationComputer VisionImage ClassificationObject DetectionOCRImage SegmentationMultimodal ModelsOpenCVPyTorchTensorFlowYOLOHugging Face VisionETLELTData LakesParquetORCMLOpsCI/CDAutomated TestingAPI DesignMicroservicesModel MonitoringData Architecture Standards & Governance

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

Own production ML outcomes
The role covers recommendations, NLP and personalization from problem definition and metric design through deployment, monitoring and post-launch improvement. Prepare a concrete account of a production ML deliverable, its evaluation and what changed after launch.
Demonstrate deployed computer vision work
Production experience with image recognition or computer vision is explicitly required, including areas such as OCR, detection, classification, segmentation or multimodal vision. Prepare to explain deployment and evaluation decisions for a relevant example.
Build reliable pipelines and data lakes
The role owns ETL/ELT pipelines that turn disparate sources into ML-ready datasets and sets standards for partitioning, schema evolution and columnar storage. Be ready to explain volumes, freshness or latency targets, backfills, failure handling, and performance or cost choices.
Engineer models as operated services
The posting emphasizes automated tests, reproducible CI/CD workflows, monitoring and versioned APIs aligned with microservices. Focus on how a deployed service was tested, integrated, monitored and improved after an incident or production finding.
Lead client-facing delivery
The role leads discovery, scope and constraint definition, technical planning, risk management and agreement on measurable success criteria. Prepare an example of translating a business need into a delivery plan and managing a scope change.

Likely challenges to think through

Balancing pipeline freshness, reliability and cost
Preparation hypothesis, not a reported employer problem: disparate sources and data-lake design may force trade-offs between freshness targets, backfill effort, query performance and storage cost. Prepare to explain a design choice using actual scale, failure modes and operating targets.
Moving quickly without treating models as prototypes
Preparation hypothesis, not a reported employer problem: delivery speed can conflict with reproducibility, testing and reliable API operation. Prepare to walk through release safeguards, monitoring signals and a post-launch adjustment for a production system.
Keeping client expectations aligned as scope changes
Preparation hypothesis, not a reported employer problem: an evolving business request may outgrow agreed constraints or success metrics. Prepare to show how you would surface the trade-off, revise the technical plan and communicate the impact on measurable outcomes.

Practice your answers

Technical practice

Suggested practice question: Defining an ML problem

What to coverHow would you turn a client’s business need into an ML problem with measurable success criteria? Preparation suggestion: distinguish business outcomes from offline model metrics, and explain how you would agree on constraints before building.

Suggested practice question: Recommendations and personalization

What to coverHow would you design and evaluate a production recommendation or personalization system? Preparation suggestion: discuss data quality, a baseline, offline evaluation, and how post-launch results would guide changes.

Suggested practice question: NLP or GenAI delivery

What to coverHow would you move an NLP or GenAI feature from prototype to a dependable client-facing service? Preparation suggestion: explain how you would evaluate output quality, test failure cases, and decide what to monitor after deployment.

Suggested practice question: Production computer vision

What to coverHow would you develop and deploy an image classification, detection, OCR, segmentation, or multimodal vision solution? Preparation suggestion: choose one task and explain its data and evaluation needs, deployment constraints, and how you would investigate errors in production.

Suggested practice question: ETL/ELT architecture

What to coverHow would you design a scalable pipeline that turns disparate sources into analytics- and ML-ready datasets? Preparation suggestion: cover transformation choices, data-quality checks, expected volumes, and the trade-off between latency and operational complexity.

Suggested practice question: Pipeline recovery

What to coverWhat would you do if a production pipeline missed its freshness target or required a historical backfill? Preparation suggestion: explain failure detection, safe reruns, validation of recovered data, and communication about affected SLAs.

Suggested practice question: Data-lake design

What to coverHow would you choose partitioning, file sizes, schema-evolution rules, and a columnar format for a data lake? Preparation suggestion: relate Parquet or ORC choices to query performance, maintainability, and storage or processing cost.

Suggested practice question: CI/CD for ML

What to coverWhat would a reproducible, CI/CD-aligned workflow look like for deploying a model? Preparation suggestion: distinguish tests for code, data, and model behavior, and explain how you would version artifacts and control releases.

Suggested practice question: Monitoring and improvement

What to coverHow would you tell whether a deployed ML system remains reliable and useful after launch? Preparation suggestion: separate service and data-health signals from model and business outcomes, then describe how evidence would trigger investigation or improvement.

Suggested practice question: Versioned model APIs

What to coverHow would you expose a model or dataset through a versioned service for integration with other applications? Preparation suggestion: discuss API contracts, compatibility when versions change, testing, and operational monitoring in a microservices setting.

Behavioural practice

Suggested practice question: Client problem discovery

What to coverTell me about a time you translated an unclear client request into a technical plan and measurable success criteria. Use situation, action, and result to show how you clarified constraints and secured stakeholder alignment.

Suggested practice question: Managing scope changes

What to coverDescribe a time a client changed scope during delivery. Use situation, action, and result to explain the trade-offs you communicated, how you managed risks and expectations, and what happened.

Suggested practice question: Owning a production issue

What to coverTell me about a data or ML production issue you took responsibility for resolving. Use situation, action, and result to cover diagnosis, coordination, restoration, and what you learned for future operations.

Suggested practice question: Communicating actionable insights

What to coverGive an example of explaining a technical finding to non-technical stakeholders and turning it into a business recommendation. Use situation, action, and result to show how you made the evidence clear and what decision followed.

Suggested practice question: Adapting while challenging assumptions

What to coverDescribe a fast-moving project where you questioned an assumption or addressed a gap without waiting for direction. Use situation, action, and result to explain how you worked with others, adapted your approach, and affected the outcome.

Sharpen your knowledge

Topics to prepare

Production ML evaluation and iteration
Posting: The role owns recommendation, NLP and personalization systems from metric design through monitoring and post-launch improvement; a production ML deliverable is a must-have. Suggested exercise: For one of these use cases, define an offline metric, a business success measure and a post-launch signal that would trigger a change.
Computer vision deployment choices
Posting: Production experience with image classification, detection, OCR, segmentation or multimodal vision is a must-have. Suggested exercise: Choose one of these tasks and explain its evaluation criteria, likely failure cases and what you would monitor after deployment.
Pipeline reliability at scale
Posting: ETL/ELT delivery must account for volumes, SLAs, latency or freshness, backfills and failures. Suggested exercise: Sketch a pipeline from disparate sources to an ML-ready dataset; specify a freshness target and how a failed run would be detected and recovered.
Data-lake design trade-offs
Posting: The role covers partitioning, schema evolution, file optimization, Parquet or ORC, and performance or cost trade-offs. Suggested exercise: Compare two partitioning approaches for a growing dataset and explain how a schema change would affect readers and backfills.
Production ML service design
Posting: CI/CD-aligned testing, reproducibility, monitoring and versioned APIs aligned with microservices architecture are required. Suggested exercise: Map a model change from validation to release, identifying what must be tested, versioned and monitored.
Client scope and measurable outcomes
Posting: The role leads discovery, technical planning, risk management and communication through scope changes. Suggested exercise: Turn an ambiguous request for personalization into clarifying questions, constraints, success criteria and a plan for communicating a scope change.

Skills to revise

Build an ML evaluation brief
Posting: Candidates must discuss evaluation and changes after launch for a production ML deliverable. Suggested exercise: Draft a one-page evaluation brief for a recommender, NLP or personalization feature, separating pre-release results from real-world outcomes and a proposed improvement.
Prototype a vision inference workflow
Posting: Hands-on production computer vision is explicitly required; listed framework examples include OpenCV, PyTorch, TensorFlow, YOLO and Hugging Face Vision models. Suggested exercise: Using a framework you know from that list, outline or implement inference for one listed vision task, then test edge cases and document deployment considerations.
Implement a recoverable ETL/ELT step
Posting: Pipelines must produce reliable analytics- and ML-ready datasets and handle backfills and failures. Suggested exercise: Build a small transformation from two input sources, add data-quality checks, then rehearse a failed run and a repeatable backfill.
Review a columnar dataset design
Posting: Practical data-lake experience includes Parquet or ORC and query-performance or cost trade-offs. Suggested exercise: Create or inspect a small Parquet or ORC dataset; propose partitioning and file-size changes, then explain their expected effects on reads and maintenance.
Specify a tested, versioned model service
Posting: Production software practices include automated testing, CI/CD, API design, monitoring and incident learnings. Suggested exercise: Define a versioned inference API contract, write test cases for inputs and failures, and outline release checks and operational alerts.
Practice a client delivery discussion
Posting: Client-facing delivery includes translating requirements into technical plans, handling scope changes and aligning stakeholders on success criteria. Suggested exercise: Role-play a discovery meeting and produce a short plan stating assumptions, delivery risks, measurable outcomes and how you would explain a changed requirement to technical and non-technical stakeholders.

Your preparation checklist

Map your evidence to the role
Prepare a one-page evidence map for the posting’s production ML, computer vision, data pipeline, data-lake, software delivery and client-facing requirements. Note the stated bachelor’s degree and 4–6 years of production experience; leave gaps explicit rather than inventing examples.
Walk through an owned ML launch
Prepare a project walkthrough showing the business problem, success metrics, model choice, deployment, monitoring, evaluation and changes after launch. The posting specifically asks for end-to-end ownership of a production ML deliverable.
Demonstrate production vision work
Prepare a separate walkthrough of a deployed computer vision or image-recognition solution, covering the task, framework, evaluation, integration and operational lessons. This is listed as must-have experience.
Diagram a data pipeline and lake design
Bring an architecture sketch with source systems, ETL/ELT stages, volumes, freshness or latency targets, backfills and failure handling. Explain partitioning, schema evolution, Parquet or ORC choices, and performance or cost trade-offs, as requested in the posting.
Rehearse a production design exercise
As interview preparation, sketch how you would serve a recommendation, NLP or GenAI capability through a versioned API. Include automated tests, reproducible releases, CI/CD, monitoring and incident response; these reflect the role’s delivery responsibilities.
Prepare a client-delivery example
Outline a case where you turned business needs into a technical plan, set measurable success criteria, managed a scope change and communicated a risk or trade-off to non-technical stakeholders. End with the business recommendation or outcome.
Draft questions for the employer
Prepare questions about the first client deliverable, expected production scale and reliability targets, how success is measured, and how responsibilities are divided with the Project Manager. These are questions to clarify the role, not assumptions about Quadratyx’s processes.

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…