New opportunity

Senior Data Scientist

FACLON LABS PRIVATE LIMITED · Mumbai, India

About this role

Faclon Labs is hiring a Senior Data Scientist in Mumbai for a full-time role focused on production-grade Industrial AI, machine learning and generative AI. The position calls for 3–5 years of experience and involves building end-to-end solutions, from data modeling and algorithm design through cloud deployment, for real-time and near-real-time industrial and IoT applications. Domains include energy management, process industries, vibration monitoring and industrial optimization.

Responsibilities

Develop and deploy time-series forecasting models using approaches such as Prophet, N-BEATS, LSTM and Transformers. Address predictive maintenance, anomaly detection, condition monitoring, vibration analysis and process optimization.
Apply optimization and control techniques, including particle swarm and other evolutionary methods, Model Predictive Control, golden batch analysis and operational optimization. Develop custom loss and cost functions aligned with business KPIs.
Build, fine-tune and integrate generative AI and LLM solutions, including RAG pipelines, prompt engineering, domain adaptation, analytics and assistants for industrial use cases. Work with agentic AI frameworks, MCP and A2A architectures.
Maintain MLOps pipelines using MLflow for experiment tracking and model registry, Airflow or Dagster for orchestration, and CI/CD for training, validation and deployment. Deploy and scale ML and LLM systems with Docker on AWS or GCP, considering reliability, performance and cost.
Influence AI architecture and modeling standards, mentor junior colleagues, and work with Product and Engineering teams to turn industrial problems into scalable solutions.

The role requires hands-on work beyond experimentation and suits someone comfortable with an agile, iterative startup environment and self-directed ownership.

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

Skills for this role

Industrial AIMachine LearningGenerative AITime-Series ForecastingProphetN-BEATSLSTMTransformersPredictive MaintenanceAnomaly DetectionCondition MonitoringVibration AnalysisProcess OptimizationParticle Swarm OptimizationEvolutionary AlgorithmsModel Predictive ControlGolden Batch AnalysisCustom Loss FunctionsLarge Language ModelsRetrieval-Augmented GenerationPrompt EngineeringDomain AdaptationAI AgentsMCPA2AMLOpsMLflowAirflowDagsterCI/CD pay/eval? none? remove accidental

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

Production-grade industrial ML
The role owns end-to-end solutions, from data modeling and algorithm design to cloud deployment, for real-time and near-real-time industrial and IoT use cases. Prepare to explain how a model becomes a reliable deployed system.
Time-series and equipment intelligence
The posting emphasizes forecasting, predictive maintenance, anomaly detection, condition monitoring, and vibration analysis. Prepare to explain model selection and validation for industrial time-series applications.
Optimization tied to operational KPIs
The role calls for optimization and control methods, including PSO, evolutionary methods, MPC, and golden batch analysis, plus custom loss and cost functions. Prepare to connect modeling choices to decision relevance and business KPIs.
Industrial GenAI integration
Responsibilities include building and integrating RAG pipelines, domain-adapted LLM solutions, industrial assistants, and emerging agentic architectures. Prepare to describe how these components would fit an industrial use case rather than treating them as standalone demos.
MLOps and senior technical influence
The posting specifies MLflow, Airflow or Dagster, CI/CD, Docker, and AWS or GCP deployment. It also expects influence on architecture and modeling standards, mentoring, and collaboration with Product and Engineering.

Likely challenges to think through

Balancing responsiveness with operating cost
Preparation hypothesis, not a confirmed employer problem: real-time industrial inference may create a trade-off among latency, reliability, model complexity, and cloud cost. Prepare a concrete approach to measuring those factors and deciding when a simpler model is preferable.
Turning predictions into useful decisions
Preparation hypothesis, not a confirmed employer problem: the most accurate forecast or anomaly detector may not produce the best operational outcome. Prepare to explain how you would define a KPI-aligned cost function, evaluate decisions, and account for constraints before applying optimization or control.
Validating domain-specific GenAI
Preparation hypothesis, not a confirmed employer problem: an industrial assistant may need useful domain responses without overstating uncertain information. Prepare a method for evaluating retrieval and responses, handling unsupported answers, and deciding what should remain under human review.

Practice your answers

Technical practice

How would you design a production ML pipeline for near-real-time industrial IoT data?

What to coverSuggested preparation: Explain how you would handle incoming time-series data, validation, inference latency and monitoring from data modeling through deployment. Discuss the trade-off between timely decisions and reliability; the posting specifies real-time and near-real-time industrial systems.

How would you choose among Prophet, N-BEATS, LSTM and Transformers for industrial forecasting?

What to coverSuggested preparation: Compare data needs, forecast horizon, interpretability, compute cost and performance against a simple baseline. Explain how you would validate on future time periods without leakage; all four approaches are named in the posting.

How would you build and evaluate a vibration-based condition-monitoring or anomaly-detection model?

What to coverSuggested preparation: Discuss signal preparation, the availability of failure labels, alert thresholds and evaluation when failures are rare. Explain how false alarms and missed faults would affect a predictive-maintenance decision.

When would you use evolutionary optimization or golden batch analysis for a process problem?

What to coverSuggested preparation: Describe how you would define operating constraints and an objective, then compare a search method such as particle swarm optimization with analysis of high-performing batches. Address the risk of recommending settings that look optimal in historical data but are not operationally feasible.

How would you formulate Model Predictive Control for industrial optimization?

What to coverSuggested preparation: Explain the roles of a predictive model, a planning horizon, constraints and repeated re-optimization. Discuss how model error and computation time affect whether the approach is practical.

How would you design a custom loss or cost function around a business KPI?

What to coverSuggested preparation: Show how you would translate a decision objective into a measurable penalty without losing sight of model accuracy. Discuss asymmetric error costs, validation against the actual KPI and unintended incentives.

How would you evaluate a RAG-based assistant for an industrial use case?

What to coverSuggested preparation: Outline retrieval, prompt design, domain adaptation and evaluation of whether answers are grounded in the retrieved material. Discuss how you would handle uncertain answers and compare improvements to retrieval with improvements to the model or prompt.

How would you decide whether an industrial LLM workflow needs agents, MCP or A2A?

What to coverSuggested preparation: Start with the task and compare a simpler workflow with an agentic one before introducing additional components. Explain what you would evaluate for tool use, coordination, reliability and operational complexity; the posting names agentic frameworks, MCP and A2A.

How would you structure experiment tracking, orchestration and release checks for an ML model?

What to coverSuggested preparation: Connect MLflow tracking and model registry to reproducible experiments, Airflow or Dagster to automated workflows, and CI/CD to training, validation and deployment checks. Explain how you would detect a failed run and avoid promoting an inadequately validated model.

How would you deploy and troubleshoot a Dockerized ML or LLM service on AWS or GCP?

What to coverSuggested preparation: Walk through a deployment design and the signals you would inspect if latency, reliability or cost deteriorated. Compare scaling and performance options against operational cost; Docker, AWS, GCP and these trade-offs are stated in the posting.

Behavioural practice

Tell me about a time you translated a domain problem into a technical solution with other teams.

What to coverSuggested preparation: Use a real situation involving stakeholders, your actions to clarify the problem and a measurable or observable result. Focus on decisions made with Product and Engineering counterparts, reflecting the collaboration described in the posting.

Describe a time you helped a less-experienced colleague improve their work.

What to coverSuggested preparation: Give the situation, the specific mentoring actions you took and the result for the colleague or team. Distinguish your contribution from the other person's work; mentoring junior team members is part of the role.

Tell me about a time you influenced a modeling or architecture standard.

What to coverSuggested preparation: Describe the existing inconsistency, how you built support for a change and what happened after adoption. Use a real example rather than claiming authority you did not have; the posting calls for influence on AI architecture and modeling practices.

Describe a time priorities changed during an iterative technical project.

What to coverSuggested preparation: Set out the situation, how you revised the plan and communicated trade-offs, and the resulting outcome. Choose an example that shows judgment under changing requirements, which is relevant to the posting's agile, iterative environment.

Tell me about a production problem you took ownership of beyond the experiment stage.

What to coverSuggested preparation: Explain the situation, your troubleshooting and coordination actions, and the result or lesson learned. Keep the example grounded in your actual responsibility; the role emphasizes self-directed ownership and production-grade delivery.

Sharpen your knowledge

Topics to prepare

Time-series forecasting choices
Posting: Forecasting approaches include Prophet, N-BEATS, LSTM and Transformers for industrial data. Revision exercise (suggestion): Explain when each approach is appropriate and how a time-ordered evaluation avoids leakage in a near-real-time setting.
Industrial condition monitoring
Posting: The role covers predictive maintenance, anomaly detection, condition monitoring and vibration analysis. Revision exercise (suggestion): Compare how you would define an anomaly, handle scarce failure examples and evaluate whether an alert is operationally useful.
Optimization and control
Posting: Responsibilities name PSO and evolutionary methods, Model Predictive Control, golden batch analysis and KPI-aligned cost functions. Revision exercise (suggestion): Sketch a process-optimization objective, its constraints and the trade-off between a search method and MPC.
Industrial GenAI architecture
Posting: The role includes RAG, prompt engineering, domain adaptation, industrial assistants, agentic frameworks, MCP and A2A architectures. Revision exercise (suggestion): Diagram a RAG-based assistant and explain where retrieval, model behavior and interactions between components need evaluation.
Production ML lifecycle
Posting: The role calls for MLflow, Airflow or Dagster, CI/CD, Docker and deployment on AWS or GCP, with reliability, performance and cost in mind. Revision exercise (suggestion): Trace a model from experiment through validation, registry and deployment, identifying what would trigger a rollback.
Senior technical collaboration
Posting: The senior contributor influences architecture and modeling standards, mentors juniors and works with Product and Engineering. Revision exercise (suggestion): Outline how to turn an industrial KPI into a model objective and explain the design and validation decisions to both technical and product colleagues.

Skills to revise

Build a forecasting comparison
Posting: The role develops and deploys time-series forecasts. Revision exercise (suggestion): On a sample industrial-style time series, compare two listed approaches using chronological validation; summarize forecast error, inference needs and your choice.
Design a monitoring alert
Posting: Anomaly detection, vibration analysis and condition monitoring are named applications. Revision exercise (suggestion): Work through a sample sensor sequence, specify an anomaly score and alert threshold, then examine false alerts and missed events.
Formulate a constrained optimization problem
Posting: The role applies PSO or evolutionary methods, MPC and custom KPI-based cost functions. Revision exercise (suggestion): Write a small process objective with operating constraints and show how you would test whether its recommended settings improve the chosen KPI.
Prototype an industrial RAG workflow
Posting: Building and integrating RAG pipelines and LLM-based assistants is part of the role. Revision exercise (suggestion): Use a small set of sample industrial documents to specify retrieval, prompting and answer checks; identify how you would detect unsupported answers.
Map an MLOps deployment pipeline
Posting: MLflow tracking and registry, Airflow or Dagster orchestration, CI/CD, Docker and AWS or GCP deployment are specified. Revision exercise (suggestion): Draft a minimal training-to-deployment workflow using those named components, including validation gates and operational monitoring.

Your preparation checklist

Prepare an industrial ML project walkthrough
The role covers forecasting, predictive maintenance and anomaly detection for industrial or IoT data. Prepare a past project, if available, or clearly label an illustrative case; show the data, modeling choices, validation, deployment path and operational outcome.
Work through an optimization example
The posting names particle swarm and evolutionary methods, Model Predictive Control, golden batch analysis and KPI-aligned cost functions. Sketch an industrial decision problem with an objective, constraints, proposed method and a way to evaluate its operational value.
Outline an industrial GenAI use case
The role includes RAG, domain adaptation, assistants and agentic architectures. Prepare a proposed use case showing what information an assistant would retrieve, how its answers would be checked, and where MCP or A2A would—or would not—fit.
Practice a production architecture design
Design a near-real-time ML deployment on paper, covering data flow, validation, monitoring and rollback. The posting specifically mentions MLflow, Airflow or Dagster, CI/CD, Docker and AWS or GCP; be ready to explain tool choices and reliability, latency and cost trade-offs.
Gather evidence of senior-level ownership
The role involves architecture standards, mentoring and collaboration with Product and Engineering. Prepare concise examples, if available, of a technical decision you owned, a colleague you supported and an industrial problem you helped turn into a deployable solution.
Write questions for the interviewers
Ask which industrial use cases are priorities, what real-time constraints and data quality challenges matter most, how deployed models are assessed, and how the Senior Data Scientist works with Product and Engineering. These are questions to clarify, not established details of the role.

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…