New opportunity

Azure AI Leader

Capgemini Invent · Bangalore, India

About this role

Capgemini Invent seeks an Azure AI Leader in Bangalore to lead the architecture, design and delivery of enterprise-grade generative and agentic AI solutions for business-critical use cases. This is a permanent position.

Responsibilities

Provide hands-on technical leadership in building AI applications, agentic workflows, RAG-based solutions, model integrations, API-enabled capabilities and reusable enterprise AI components.
Architect solutions using Microsoft Copilot Studio and Azure AI Foundry, including custom agents, prompt engineering, workflow orchestration and code-first patterns using Azure AI Foundry Agent Service.
Design and govern multi-agent architectures using connected or child agents, agent-to-agent (A2A) collaboration and Model Context Protocol (MCP) to integrate tools, workflows and enterprise data.

Profile:

Bring deep expertise in generative AI solution architecture, LLM-based application design, AI platform engineering and enterprise-scale implementation.
Understand enterprise data grounding, RAG architectures, vector databases, knowledge indexing, API integration and secure connectivity to enterprise systems.
Have hands-on experience with conversational AI, task-oriented agents, autonomous or semi-autonomous workflows, and enterprise copilots or assistants. Practical experience with Azure AI Foundry across building, deploying, observing, governing and managing enterprise AI agents is required.

The posting mentions flexible work arrangements, including remote work or flexible hours, along with career-growth programs and opportunities to earn certifications in technologies such as generative AI. It does not specify a role-specific remote arrangement or an experience minimum.

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

Skills for this role

Generative AIAgentic AILLMsAI solution architectureAI platform engineeringRAGVector databasesKnowledge indexingMicrosoft Copilot StudioAzure AI FoundryAzure AI Foundry Agent ServicePrompt engineeringAgent orchestrationMulti-agent architecturesAgent-to-agent (A2A) collaborationModel Context Protocol (MCP)API integrationConversational AIEnterprise data groundingAI model integration

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

End-to-end enterprise AI delivery
The role leads architecture, design and delivery of generative and agentic AI solutions for business-critical use cases. Prepare to explain how you would carry an architecture through implementation and enterprise-scale delivery.
Hands-on, reusable AI applications
The posting calls for technical leadership in building AI applications, workflows, model integrations, API-enabled capabilities and reusable components. Prepare to discuss design decisions at both application and shared-component level.
Grounded AI with enterprise connectivity
RAG, vector databases, knowledge indexing, API integration and secure connectivity to enterprise systems are explicit requirements. Prepare to explain how retrieval and data access support a grounded response.
Azure agent implementation and governance
The role specifies Copilot Studio, Azure AI Foundry and code-first Agent Service patterns, plus practical Foundry experience across building, deploying, observing and governing agents. It also calls for multi-agent designs using connected or child agents, A2A and MCP.

Likely challenges to think through

Balancing agent autonomy with control
A likely design challenge, not a stated employer problem, is deciding when autonomous or multi-agent workflows justify added coordination and governance. Prepare a decision framework comparing a simpler task-oriented agent with connected agents, including how behavior would be observed.
Choosing an agent implementation pattern
Because the role spans Copilot Studio, Foundry and code-first Agent Service patterns, a likely trade-off is choosing an approach without adding unnecessary implementation complexity. Prepare to justify a choice against workflow needs, integration requirements and maintainability.
Grounding across protected enterprise data
The stated combination of RAG and secure enterprise connectivity suggests a likely tension between useful retrieval and controlled data access. Prepare to explain how indexing, retrieval and API access would fit together while respecting system boundaries.

Practice your answers

Technical practice

How would you architect an enterprise-grade generative AI solution for a business-critical use case?

What to coverThe role covers end-to-end architecture, design and delivery. Preparation suggestion: explain how you would define the use case, choose application components, connect enterprise data and evaluate whether the solution meets its intended purpose.

How would you design and evaluate a RAG pipeline grounded in enterprise data?

What to coverThe posting names RAG, knowledge indexing, vector databases and enterprise data grounding. Preparation suggestion: discuss indexing and retrieval choices, how you would check answer grounding, and how you would diagnose poor retrieval versus poor generation.

When would you use Copilot Studio versus Azure AI Foundry for a custom agent?

What to coverBoth platforms are named in the role, alongside custom agent development. Preparation suggestion: compare their fit for a given agent design, explain any integration trade-offs, and identify what you would validate before choosing.

How would you implement a code-first agent with Azure AI Foundry Agent Service?

What to coverThe role explicitly mentions Agent Service-based code-first patterns and workflow orchestration. Preparation suggestion: describe the agent's instructions, tool interactions and workflow boundaries, then explain how you would test its behavior.

How would you choose and integrate a model for an enterprise AI application?

What to coverModel integration and LLM-based application design are stated responsibilities. Preparation suggestion: explain how you would assess model output, prompt behavior and application constraints, including when a different model or prompt design might help.

How would you design a multi-agent workflow using connected or child agents and A2A collaboration?

What to coverThe role calls for designing and governing these multi-agent patterns. Preparation suggestion: define agent responsibilities, handoffs and failure handling, and explain when multiple agents add value over a simpler workflow.

How would you use MCP to connect agents to enterprise tools and data?

What to coverThe posting names MCP for standardized tool, workflow and enterprise data integration. Preparation suggestion: explain the integration boundary, how an agent would select and use tools, and how you would limit inappropriate access.

How would you build secure, API-enabled AI capabilities that can be reused?

What to coverAPI integration, secure enterprise connectivity and reusable AI components are all in scope. Preparation suggestion: describe component interfaces, access boundaries and error handling, including how you would avoid duplicating integration logic.

How would you deploy, observe and govern an agent built with Azure AI Foundry?

What to coverPractical experience across building, deploying, observing, governing and managing agents is specified. Preparation suggestion: discuss what behavior and failures you would monitor, how you would evaluate changes, and how findings would inform ongoing management.

When should an enterprise workflow use an autonomous rather than a semi-autonomous agent?

What to coverThe profile mentions both autonomous and semi-autonomous workflows, as well as task-oriented agents. Preparation suggestion: compare the value of independent action with the need for oversight, and explain how you would set action boundaries for the task.

Behavioural practice

Tell me about a time you led a complex AI solution from design through delivery.

What to coverThe role calls for end-to-end technical leadership. Preparation suggestion: use a real situation/action/result example that makes your architectural decisions, hands-on contribution and delivery outcome clear.

Describe a time you worked with stakeholders to shape a business-critical AI use case.

What to coverThe posting describes collaboration with clients and solutions tailored to their challenges. Preparation suggestion: give a real situation/action/result example showing how you clarified the need, handled differing priorities and established what success meant.

Tell me about a time you resolved a difficult issue in an AI application or workflow.

What to coverThe role includes hands-on building and delivery of AI applications and agentic workflows. Preparation suggestion: use a real situation/action/result example to show how you isolated the issue, weighed remedies and verified the result.

Describe a time you balanced an AI agent's capabilities with appropriate governance.

What to coverDesigning and governing multi-agent architectures is an explicit responsibility. Preparation suggestion: give a real situation/action/result example explaining the trade-off, the boundaries you established and the observed outcome.

Tell me about a time you made an AI component reusable across multiple needs.

What to coverThe role includes building reusable enterprise AI components. Preparation suggestion: use a real situation/action/result example covering why reuse mattered, how you designed for it and what happened when others used the component.

Sharpen your knowledge

Topics to prepare

Enterprise GenAI architecture
The role leads end-to-end architecture and delivery of business-critical GenAI and agentic AI solutions. Suggested exercise: sketch an application architecture and explain its model integration, API boundaries, reusable components and delivery trade-offs.
Enterprise grounding and RAG
The posting explicitly names data grounding, RAG, vector databases and knowledge indexing. Suggested exercise: diagram how an enterprise document moves through indexing, retrieval and grounding, then explain how you would assess an unsupported answer.
Multi-agent design and governance
Connected or child agents, A2A collaboration and MCP are named responsibilities. Suggested exercise: map agent responsibilities and handoffs for a task-oriented workflow, identifying where tools and enterprise data are exposed through MCP.
Copilot Studio and Azure AI Foundry patterns
The role calls for solutions using both platforms, including custom agents and code-first patterns with Azure AI Foundry Agent Service. Suggested exercise: compare where you would use each named approach in one enterprise-assistant design and justify the choice.
Agent lifecycle at enterprise scale
The posting requires practical Azure AI Foundry experience across building, deploying, observing, governing and managing agents. Suggested exercise: outline the decisions and evidence you would review at each stage for a business-critical agent.

Skills to revise

Build and observe a Foundry agent
Suggested exercise: build a small agent in Azure AI Foundry, deploy it in a practice setting, inspect its observed behaviour and document a governance decision. This rehearses the hands-on Foundry lifecycle explicitly required.
Create a Copilot Studio custom agent
Suggested exercise: configure a task-oriented custom agent in Microsoft Copilot Studio and walk through its conversation and workflow. The posting calls for Copilot Studio implementation and enterprise assistants.
Implement a grounded RAG flow
Suggested exercise: index a small set of sample documents, retrieve relevant passages and use them to ground a model response; check whether the response is supported. This practises the stated RAG, indexing and model-integration work.
Prototype code-first orchestration
Suggested exercise: use Azure AI Foundry Agent Service to prototype a code-first agent workflow, then trace its steps and handoffs. Code-first agent patterns and workflow orchestration are explicit responsibilities.
Connect agents to enterprise-style tools
Suggested exercise: with mock data and APIs, prototype an MCP-connected tool and an A2A handoff between agents; document the access boundaries. The role names MCP, A2A, API integration and secure enterprise connectivity.
Refine prompts for a task-oriented assistant
Suggested exercise: write and revise prompts for an assistant that must use grounded information and hand off a task; compare its responses on a small set of test requests. Prompt engineering and task-oriented agents are both named in the posting.

Your preparation checklist

Prepare an end-to-end project walkthrough
If you have a relevant example, make a one-page account of the business use case, your architecture and hands-on leadership decisions, delivery challenges, and outcomes. The role calls for leading enterprise-grade generative and agentic AI delivery.
Sketch an enterprise RAG design
As interview practice, draw a design showing data grounding, knowledge indexing, vector retrieval, model integration, API access, and secure connectivity. Be ready to explain the trade-offs; these are all named in the posting.
Gather Azure AI Foundry evidence
Prepare specific examples, where available, of building, deploying, observing, governing, and managing agents in Azure AI Foundry. Distinguish what you implemented yourself from what you directed.
Compare agent implementation approaches
Draft a short design exercise showing when you would use Copilot Studio, a custom agent, or a code-first Azure AI Foundry Agent Service pattern, including prompt and workflow orchestration choices.
Explain a multi-agent integration design
Sketch how connected or child agents could collaborate using A2A and access tools or enterprise data through MCP. Identify boundaries and governance decisions you would discuss; the sketch is preparation, not a claimed employer architecture.
Bring focused questions for the employer
Ask which business-critical use cases take priority, how delivery and agent governance responsibilities are divided, and which work arrangement applies to this Bangalore role. The posting discusses flexibility generally but does not specify a role-specific arrangement.

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…