Role Purpose: Kotak Neo is seeking a Senior AI & ML Engineer to join their team in Mumbai or Bangalore. The role involves designing and building agentic systems, retrieval pipelines, and applied machine learning solutions to enhance their trading platform and customer experience. Responsibilities: Design and build agentic systems and retrieval pipelines for generative AI, including tool/function calling, multi-step execution, and state management. Develop applied machine learning solutions for personalization, client intelligence, risk signals, and anomaly detection. Architect ML platform components such as feature infrastructure, training pipelines, and real-time/batch serving. Implement rigorous engineering practices including testing, CI/CD, and regression suites for prompt engineering. Collaborate cross-functionally to ensure system stability and regulatory compliance. Evaluate and optimize system performance using defined metrics. Qualifications: Minimum 5 years of software engineering experience, with at least 3 years specifically in machine learning, data, or AI systems. Strong proficiency in Python and experience with cloud infrastructure, containerized deployments, and orchestration tools. Data fluency including SQL and experience managing large datasets with Spark or equivalent technologies. Deep expertise in one of the following tracks: generative AI, applied machine learning, or ML platform, with working familiarity across the others. A measurement-led mindset with a track record of taking systems to production with dedicated ownership.
For full details about this role, click “Apply on company site” to view the employer’s original posting.
A focused preparation guide, built around this job’s responsibilities and requirements.
✦ AI-generated guide
Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.
01
Understand the role
What matters most
Building Agentic Generative AI Systems
The role requires designing and implementing complex agentic workflows, including multi-step execution, tool calling, and state management. You must demonstrate how to build robust retrieval pipelines that go beyond simple RAG to handle sophisticated user queries within a trading platform context.
Production-Grade ML Engineering
You are expected to own the end-to-end lifecycle of ML solutions, from stakeholder framing to deployment. This includes architecting scalable feature infrastructure, training pipelines, and real-time serving, while maintaining rigorous engineering standards like CI/CD and automated regression testing.
Measurement-Led Performance Optimization
Success is defined by your ability to evaluate system performance against clear metrics. You must be prepared to discuss how you monitor for model drift, feature decay, and prompt engineering regressions to ensure stability in a high-stakes financial environment.
Cross-Functional Regulatory Compliance
Given the financial domain, you must balance innovation with strict adherence to regulatory requirements. You should be ready to explain how you design systems that are not only performant but also deterministic, auditable, and stable enough for a retail trading platform.
Likely challenges to think through
Balancing Determinism with Generative AI
Generative AI models are inherently probabilistic, which conflicts with the deterministic requirements of financial trading systems. Prepare to discuss strategies for implementing guardrails, fallback mechanisms, and validation layers that ensure AI outputs remain accurate and safe for users.
Managing High-Volume Financial Data
The platform serves 5 million customers, requiring efficient handling of large datasets via Spark or similar tools. You may face challenges in optimizing feature stores for low-latency real-time inference while ensuring data consistency across batch and streaming pipelines.
System Stability in a Legacy-Modern Hybrid
Kotak Neo operates on a modern stack built upon a legacy broking infrastructure. You will likely encounter technical debt or integration complexities; prepare to discuss how you maintain system reliability and performance when bridging new AI features with established, high-volume trading backends.
03
Practice your answers
Technical practice
Designing Agentic Systems
What to coverExplain your approach to implementing multi-step execution and state management in agentic workflows. Discuss how you handle tool/function calling and implement fallback mechanisms to ensure reliability when an agent fails to reach a goal.
Retrieval Pipelines for Generative AI
What to coverDescribe the architecture of a robust retrieval pipeline. Focus on how you manage data ingestion, indexing, and the trade-offs between different retrieval strategies to improve the relevance of LLM-generated responses.
ML Platform Architecture
What to coverDetail the components required for a scalable ML platform, specifically feature infrastructure and training pipelines. Explain how you ensure reproducibility in training and manage the transition from batch to real-time serving.
Monitoring and Drift Detection
What to coverHow do you architect monitoring for production ML models to detect feature decay and concept drift? Suggest specific metrics and automated alerting strategies to maintain system performance over time.
Prompt Engineering and Regression Testing
What to coverSince the role requires rigorous engineering for prompt-based systems, explain how you build regression suites for prompts. Discuss how you measure the impact of prompt changes on output quality and consistency.
Data Engineering with Spark
What to coverDescribe your experience managing large datasets using Spark. Focus on how you optimize data processing jobs for performance and cost-efficiency in a cloud-based environment.
Anomaly Detection in Financial Systems
What to coverDiscuss the challenges of building anomaly detection models for trading platforms. Address how you balance sensitivity and specificity to minimize false positives while maintaining regulatory compliance.
CI/CD for Machine Learning
What to coverExplain how you integrate ML workflows into a standard CI/CD pipeline. Address the specific challenges of versioning models, data, and code simultaneously to ensure reliable deployments.
Behavioural practice
Taking Ownership of Production Systems
What to coverDescribe a situation where you took full ownership of an ML system from development to production. Focus on how you handled on-call responsibilities and resolved a critical production issue.
Cross-Functional Collaboration
What to coverProvide an example of working with non-technical stakeholders to frame an ML problem. Explain how you translated business requirements into technical metrics and ensured the final solution met their needs.
Measurement-Led Decision Making
What to coverShare an experience where you had to evaluate a system's performance using defined metrics. Explain how you used these metrics to justify a design change or an optimization effort.
Navigating Regulatory Constraints
What to coverDescribe a time you had to build a technical solution that required strict adherence to regulatory or compliance standards. How did you ensure your design remained stable and compliant without sacrificing performance?
02
Sharpen your knowledge
Topics to prepare
Agentic System Architecture
Focus on the design patterns for multi-step execution, state management, and tool/function calling in generative AI. Exercise: Draft a high-level architecture diagram for an AI agent that retrieves real-time stock data to answer complex user queries, specifically detailing how you would handle state persistence and fallback mechanisms when a tool call fails.
ML Platform Engineering
Understand the lifecycle of production ML, specifically feature infrastructure and monitoring for drift. Exercise: Design a monitoring strategy for a risk signal model, identifying which specific metrics you would track to detect feature decay and how you would trigger automated retraining pipelines.
Regression Testing for Prompt Engineering
Explore methodologies for ensuring deterministic behavior in non-deterministic LLM outputs. Exercise: Propose a framework for a regression suite that validates prompt changes against a golden dataset of inputs and expected outputs to ensure system stability before deployment.
Skills to revise
Production-Grade Python
Demonstrate proficiency in writing testable, maintainable code suitable for high-stakes trading environments. Exercise: Refactor a script into a modular, unit-tested class structure that includes error handling and logging, simulating a production service component.
Large-Scale Data Processing with Spark
Focus on data fluency and managing large datasets for training or feature engineering. Exercise: Write a Spark job to aggregate historical trading data into daily feature vectors, ensuring the process is optimized for memory usage and handles potential data skew.
CI/CD for ML Systems
Implement automated pipelines for model deployment and infrastructure management. Exercise: Outline a CI/CD pipeline configuration that includes automated linting, unit testing for data validation, and a deployment step to a containerized environment using Docker and an orchestration tool.
04
Your preparation checklist
Prepare a production-grade ML system case study
Select a past project where you owned the end-to-end lifecycle. Prepare to discuss how you handled feature infrastructure, training pipelines, and real-time serving, specifically highlighting how you measured performance and managed model drift or feature decay.
Develop a technical walkthrough for agentic systems
Create a brief architectural diagram or explanation of an agentic system you have built or designed. Focus on your approach to tool/function calling, state management, and implementing fallback mechanisms to ensure system reliability.
Review MLOps and engineering best practices
Be ready to explain your experience with CI/CD for ML, specifically how you implement regression testing for prompt engineering. Prepare examples of how you have integrated testing and version control into your machine learning workflows.
Formulate questions on regulatory and stability constraints
Prepare thoughtful questions regarding how Kotak Neo balances rapid AI innovation with the regulatory requirements of the financial sector. Ask about their current approach to monitoring system stability and deterministic performance in production.
Synthesize data engineering experience with Spark
Prepare concrete examples of how you have managed large datasets using Spark or equivalent technologies. Be ready to discuss how you optimized data pipelines to support personalization or risk signal models.
Availability has not yet been verified. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy
Share this opportunityKnow someone who would be a great fit?