New opportunity

Software Development Engineer II (ML), CMT

Amazon · Bengaluru/Chennai, India

About this role

The Competitive Monitoring team at Amazon is seeking a Software Development Engineer II (ML) to build intelligence engines that power competitive advantage. You will architect real-time systems that ingest, analyze, and act on competitive data at scale using machine learning and agentic workflows. Responsibilities include designing autonomous agentic systems, productionizing and maintaining ML models, monitoring model health and data drift, and building observability and alerting infrastructure. You will own the full stack, from APIs to backend systems, ensuring performance, security, and scalability. Key Qualifications: 3+ years of professional software development experience, proficiency in at least one programming language, and a Bachelor's degree in computer science or equivalent. Preferred Qualifications: 4+ years of experience building large-scale ML infrastructure for recommendation, ads ranking, personalization, or search; deep knowledge of ML and LLM fundamentals including transformer architecture, training/inference lifecycles, and optimization techniques.

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

Skills for this role

Machine LearningLarge Language ModelsAgentic WorkflowsDistributed SystemsSoftware DevelopmentMLOpsObservabilitySystem ArchitectureData Drift MonitoringTransformer Architecture

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

Architecting Autonomous Agentic Workflows
The role requires designing systems that can autonomously solve complex business problems. You must demonstrate the ability to structure agentic workflows that integrate LLMs into production environments, ensuring they operate reliably and safely at scale.
Full-Stack ML Infrastructure Ownership
You are expected to own the entire stack, from APIs and backend systems to the deployment and maintenance of ML models. This requires balancing traditional software engineering best practices with the specific demands of MLOps, such as model versioning and inference optimization.
Observability and Model Health Monitoring
A core priority is building infrastructure to monitor model performance and detect data drift. You must be prepared to design systems that provide real-time visibility into AI outputs and trigger alerts when model reliability degrades.

Likely challenges to think through

Balancing Autonomy with System Reliability
Agentic systems can be unpredictable. The challenge lies in building guardrails that allow for autonomous decision-making while ensuring the system remains resilient and observable. Prepare by discussing strategies for implementing circuit breakers, human-in-the-loop verification, and robust error handling for non-deterministic AI outputs.
Scaling Real-Time Competitive Intelligence
Ingesting and analyzing competitive data at Amazon's scale introduces significant latency and throughput constraints. You will need to demonstrate how to optimize inference lifecycles and data pipelines to ensure real-time performance without compromising system availability.

Practice your answers

Technical practice

Designing Agentic Workflows

What to coverExplain how you would architect an autonomous agentic system to process competitive data. Focus on the orchestration of LLM calls, state management, and how you ensure the system remains within safety and performance bounds.

Detecting and Mitigating Data Drift

What to coverDescribe your approach to monitoring model health in a production environment. Discuss specific metrics or statistical methods you would use to identify data drift and the automated strategies you would implement to trigger retraining or alerts.

Optimizing Transformer Models for Inference

What to coverDiscuss techniques for optimizing transformer-based models for real-time inference at scale. Address trade-offs between latency, throughput, and model accuracy, such as quantization, pruning, or knowledge distillation.

Building Observability for AI Systems

What to coverHow do you design observability infrastructure for non-deterministic AI outputs? Explain how you would implement tracing and logging to debug agentic decision-making processes in a distributed system.

Scaling Distributed ML Infrastructure

What to coverDescribe the challenges of deploying ML models that handle massive volumes of competitive data. Focus on how you would design the backend architecture to ensure high availability and low latency for API-driven requests.

Productionizing ML Models

What to coverDetail your process for moving a model from a research environment to a production-ready service. Suggestion: Focus on CI/CD pipelines for ML, versioning, and automated testing strategies for model artifacts.

Handling Real-Time Data Ingestion

What to coverExplain the architectural patterns you would use to ingest and process high-velocity competitive data. Discuss how you would ensure data consistency and system resilience during peak traffic periods.

LLM Lifecycle Management

What to coverDescribe the end-to-end lifecycle of an LLM-based application, from fine-tuning or RAG implementation to deployment and monitoring. Address how you manage model updates without disrupting existing services.

Behavioural practice

Collaborating Across Global Teams

What to coverDescribe a situation where you had to align technical goals with a team in a different geographic location. Focus on how you communicated complex architectural decisions and resolved disagreements to ensure project success.

Ownership of Full-Stack Systems

What to coverProvide an example of a time you took end-to-end ownership of a complex system. Explain how you balanced the need for rapid feature delivery with the requirement for long-term maintainability and performance.

Handling Ambiguity in Technical Projects

What to coverTell me about a time you were tasked with solving a problem with unclear requirements or evolving technology. Explain the steps you took to define the scope and the actions you took to deliver a robust solution.

Prioritizing Technical Debt vs. Innovation

What to coverDescribe a scenario where you had to choose between implementing a new feature and addressing technical debt. Explain your decision-making process and the impact your choice had on the team's velocity and system stability.

Sharpen your knowledge

Topics to prepare

Agentic Workflow Architecture
The role requires designing autonomous systems that use LLMs to solve complex business problems. Revision exercise: Draft a high-level design for an agentic system that monitors competitor pricing, including decision-making loops, tool-use capabilities, and safety guardrails. This is directly relevant to the team's goal of building intelligence engines for competitive monitoring.
ML Model Observability and Drift
The position emphasizes maintaining model health and detecting data drift in production. Revision exercise: Define a strategy for monitoring model performance metrics and input data distributions in real-time. Identify specific triggers for alerting and automated retraining. This is critical for the stated responsibility of ensuring reliability in large-scale ML infrastructure.

Skills to revise

Productionizing ML Models
You must bridge the gap between model development and deployment. Revision exercise: Create a CI/CD pipeline outline for an ML model, focusing on versioning, automated testing, and deployment strategies like canary releases. This addresses the requirement to accelerate the ML/AI path to production.
Transformer Architecture Optimization
The role requires deep knowledge of LLM fundamentals and optimization. Revision exercise: Review techniques for optimizing transformer inference, such as quantization, pruning, or knowledge distillation. This is relevant to the preferred qualification regarding training/inference lifecycles and performance at scale.
Distributed Systems Design
The role involves building systems that handle massive volumes of data. Revision exercise: Practice designing a distributed ingestion service that can process high-throughput competitive data streams while maintaining low latency. This is essential for the requirement to build scalable, robust solutions at Amazon scale.

Your preparation checklist

Architect a scalable agentic workflow
Prepare a system design walkthrough for an autonomous agentic system that processes high-volume data. Focus on how you would handle real-time ingestion, maintain observability, and ensure the system remains resilient and secure at scale.
Demonstrate MLOps and production lifecycle expertise
Prepare to discuss your experience in productionizing ML models. Specifically, outline your approach to monitoring model health, detecting data drift, and automating the path from model development to deployment.
Review transformer architecture and LLM fundamentals
Refresh your knowledge on transformer architectures, training and inference lifecycles, and optimization techniques. Be ready to explain how these concepts apply to large-scale recommendation, search, or personalization systems.
Prepare examples of full-stack ownership
Identify specific projects where you owned the end-to-end stack, from designing APIs to managing backend infrastructure. Highlight how you balanced performance, scalability, and usability in those implementations.
Formulate questions on competitive intelligence challenges
Prepare thoughtful questions regarding the specific technical hurdles the Competitive Monitoring team faces, such as how they prioritize data collection or how they measure the impact of agentic workflows on customer shopping decisions.

Availability last checked: 18 Oct 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…