New opportunity

AI / ML Engineer

Accenture · Pune, India

About this role

Accenture is hiring an AI / ML Engineer for its Pune office. The full-time role focuses on developing production-ready AI applications and systems using AI tools and cloud AI services, with cloud or on-premises application pipelines. Generative AI models are a core part of the solutions; work may also involve deep learning, neural networks, chatbots and image processing.

Responsibilities and technical requirements:

Develop, integrate, deploy and optimize AI applications, particularly cloud-based solutions built at scale with Azure Cloud Services, Azure OpenAI and Azure AI Search.
Apply advanced architecture patterns including Retrieval-Augmented Generation (RAG), Agentic RAG, MCP, function calling and A2A, especially in Azure environments.
Work with prompt engineering, context engineering, vector databases, embedding and chunking strategies, and real-time data integration. Evaluate model outputs and optimize performance.
Bring experience in Java and Python and proficiency with AI/ML frameworks such as Spring AI, LangChain, LangGraph, LlamaIndex, Agno, Pydantic and TensorFlow.

Qualifications

The posting lists 5–10 years of experience and requires a minimum of 5 years in Generative AI, plus 15 years of full-time education. Generative AI is the required skill; no additional preferred skills are specified.

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

Skills for this role

Generative AIPythonJavaSpring AILangChainLangGraphLlamaIndexAgnoPydanticTensorFlowAzure Cloud ServicesAzure OpenAIAzure AI SearchRetrieval-Augmented Generation (RAG)Agentic RAGMCPFunction callingA2APrompt engineeringContext engineeringVector databasesEmbedding strategiesChunking strategiesReal-time data integrationModel evaluationModel optimizationCloud AI integrationAI application deployment

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-ready AI applications
The role develops and deploys AI systems with production-ready cloud or on-premises pipelines. Prepare to explain how you would take a GenAI application from model integration through deployment and optimization, using Java or Python where relevant.
Azure AI delivery at scale
The posting specifically calls for hands-on work with Azure Cloud Services, Azure OpenAI and Azure AI Search. Prepare an architecture walkthrough that connects these services to an application and explains how it would be deployed at scale.
Advanced GenAI architecture
RAG, Agentic RAG, MCP, function calling and A2A are named patterns, especially in Azure environments. Prepare to explain when each pattern is useful and how it changes application design.
Retrieval and context quality
The role calls out prompt and context engineering, vector databases, embedding and chunking strategies, and real-time data integration. Prepare to discuss how those choices affect what information a model receives.
Evaluation and optimization
Evaluating model output and optimizing performance are explicit expectations. Prepare a concrete approach to assessing response quality and using the findings to improve the application.

Likely challenges to think through

Freshness versus retrieval quality
Preparation hypothesis, not a stated employer problem: real-time data integration may make it harder to keep retrieved context both current and relevant. Be ready to explain how you would test embedding, chunking and retrieval choices against those competing needs.
Agent flexibility versus control
Preparation hypothesis, not a stated employer problem: agentic RAG and tool-oriented patterns can handle more complex tasks but add orchestration and failure paths. Prepare to contrast a simpler RAG flow with an agentic one and explain how you would evaluate their outputs.
Quality versus deployment performance
Preparation hypothesis, not a stated employer problem: improving model output may add steps to a production pipeline. Prepare to describe how you would measure whether a prompt, retrieval or architecture change improves quality enough to justify its performance impact.

Practice your answers

Technical practice

Suggested practice: How would you take a GenAI application from development to a production-ready cloud or on-premises pipeline?

What to coverThe role calls for production-ready AI applications and cloud or on-premises pipelines. Discuss integration, deployment, monitoring, and how you would diagnose failures after release.

Suggested practice: How would you build a scalable AI application using Azure OpenAI and Azure AI Search?

What to coverBoth services are named in the posting. Explain their roles in the application and the trade-offs you would evaluate when scaling deployment.

Suggested practice: How would you design and troubleshoot a Retrieval-Augmented Generation system?

What to coverRAG is an explicitly named architecture pattern. Cover retrieval, grounding the generated answer in retrieved content, and how you would investigate irrelevant results or unsupported answers.

Suggested practice: When would you choose Agentic RAG over a simpler RAG flow?

What to coverAgentic RAG is specifically listed. Compare the value of multi-step decisions with added complexity, latency, and the need to assess intermediate steps.

Suggested practice: How would you decide whether to use function calling, MCP, or A2A in an AI application?

What to coverThe posting names all three patterns. Describe the interaction each could support, then explain how you would choose and validate a design without assuming every application needs them.

Suggested practice: How would you choose embeddings, chunking, and vector-database strategies?

What to coverAll three are stated areas of experience. Explain how document structure and retrieval quality inform your choices, and how you would test changes rather than relying on a single setting.

Suggested practice: How would you improve an application through prompt and context engineering?

What to coverThe posting explicitly calls for both. Distinguish instruction design from selecting and organizing context, and describe how you would check whether a change improves output.

Suggested practice: How would you integrate real-time data into a GenAI application?

What to coverReal-time data integration is named in the role. Discuss data freshness, failure handling, and how you would prevent outdated or unavailable inputs from producing misleading responses.

Suggested practice: How would you evaluate model output and decide what to optimize?

What to coverThe posting calls for output evaluation and optimization. Propose task-relevant quality checks and explain how you would balance improvements against latency and resource use.

Suggested practice: How would you select between Java and Python and the listed AI/ML frameworks for an implementation?

What to coverThe posting names both languages and frameworks including Spring AI, LangChain, LangGraph, LlamaIndex, Agno, Pydantic, and TensorFlow. Compare choices against the application’s integration and deployment needs without suggesting that every framework must be used.

Behavioural practice

Suggested practice: Tell me about a time you moved an AI application toward production-ready quality.

What to coverThe role emphasizes production-ready delivery. Use a real situation, your actions on deployment or reliability, and the observable result; do not assume a particular cloud environment if it was not part of your example.

Suggested practice: Describe a time you chose between competing AI architecture approaches.

What to coverThe posting lists several GenAI patterns. Give the situation, the criteria and trade-offs behind your decision, and what the result showed.

Suggested practice: Tell me about a time model-output evaluation changed your implementation.

What to coverEvaluating and optimizing output are stated duties. Explain the situation, what evidence prompted your action, and the measured or observed result.

Suggested practice: Describe a difficult integration or deployment problem you resolved in an AI application.

What to coverThe role includes AI integration and deployment. Set out the failure or constraint, how you isolated and addressed it, and what changed afterward.

Sharpen your knowledge

Topics to prepare

Production-ready GenAI architecture
Sketch a cloud and an on-premises application pipeline, then explain where model integration, deployment, evaluation and optimization fit. This reflects the role’s focus on production-ready AI applications and systems.
RAG and retrieval design
Compare two chunking and embedding strategies for a document collection, explaining how retrieval quality, context selection and Azure AI Search affect generated answers. The posting explicitly names RAG, vector databases, embeddings, chunking and context engineering.
Agentic interaction patterns
Draw an Agentic RAG workflow and explain when function calling, MCP or A2A would be relevant, including a failure case at each interaction boundary. These are architecture patterns named in the posting.
Azure AI integration at scale
Outline how an application would use Azure OpenAI and Azure AI Search together, and identify what you would measure before and after an optimization. The posting emphasizes Azure-based AI integration, deployment and scale.
Model-output evaluation
Define a small test set for a retrieval-backed application and specify how you would inspect answer quality and retrieval failures. This prepares for the stated work of evaluating model output and optimizing performance.

Skills to revise

Implement a retrieval-backed prototype
As a preparation exercise, build a small Python or Java flow that chunks sample documents, retrieves relevant passages and supplies them as model context. This practices the posting’s RAG, embedding/chunking and AI application development requirements; the prototype is a suggestion, not a specified assignment.
Practice Azure service integration
Prepare to walk through code or pseudocode connecting an application to Azure OpenAI and Azure AI Search, including request, retrieval and response handling. Hands-on use of both services is explicitly listed.
Work through framework trade-offs
Implement or trace one application flow in a framework you know from the posting’s list—such as Spring AI, LangChain or LangGraph—and explain why you chose it over another listed option. The posting asks for Java and Python experience and names these frameworks as examples.
Test prompts and outputs
Revise a prompt and its supplied context against a fixed set of queries; record output changes, failure cases and a proposed improvement. This directly exercises the stated prompt engineering, context engineering, model evaluation and optimization work.

Your preparation checklist

Map your evidence to the stated requirements
Prepare a one-page record of dates, roles and examples relevant to the posting’s 5–10 years of experience, minimum five years in Generative AI, and 15 years of full-time education. Keep the evidence factual.
Prepare a production AI project walkthrough
Choose a project you can discuss and outline its purpose, your contribution, architecture, cloud or on-premises pipeline, deployment, and optimization. The posting emphasizes production-ready AI applications; use measured results where available.
Sketch an Azure-based AI application
As interview practice, draw a scalable design using Azure OpenAI and Azure AI Search, then explain integration, deployment and operational trade-offs. These services are named in the posting; the design is a preparation exercise, not a stated interview task.
Explain a retrieval and agent design
Prepare a diagram showing how you would choose embeddings, chunking, vector search, prompts and context for RAG. Be ready to distinguish when Agentic RAG, MCP, function calling or A2A would add value rather than using each by default.
Collect implementation and evaluation examples
Select examples that show your work in Java or Python and relevant listed frameworks, such as Spring AI or LangChain. Prepare to explain how you evaluated model outputs, handled real-time data where applicable, and used findings to optimize performance.
Draft focused questions for Accenture
Ask which Azure services and architecture patterns the Pune team uses, what production-quality and model-evaluation expectations apply, and how responsibilities are divided within the advertised Team Lead/Consultant role. These are questions to clarify, not established details of the job.

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…