New opportunity

Gen AI Engineer

Infosys Limited · Bangalore, India

About this role

Infosys Limited seeks a Gen AI Engineer in Bangalore for its Data & Analytics Unit. The role focuses on building and operating enterprise Generative AI applications using large language models (LLMs).

Responsibilities

Design, develop, deploy and maintain LLM applications and pipelines across development, testing and production environments.
Develop retrieval-augmented generation (RAG) solutions using enterprise knowledge repositories, and integrate LLMs with APIs, databases, business applications and cloud services.
Apply LLMOps practices for deployment, monitoring, evaluation, governance and lifecycle management. Monitor performance, latency, costs, hallucinations and model drift; automate deployment and versioning through CI/CD pipelines.
Work with data scientists, ML engineers and cloud teams on scalable solutions while maintaining security, compliance, observability and responsible AI practices.

Requirements

Bachelor of Engineering or integrated Bachelor of Technology. The posting lists 3–5 years of work experience; its additional requirements separately state 2–5 years in AI/ML, Generative AI, LLM engineering or MLOps.
Understanding of LLM lifecycle management and production deployment, plus experience building enterprise AI assistants, chatbots, search or knowledge-management solutions. Problem-solving, analytical, communication and collaboration skills are requested.
Listed technical requirements include Docker, Kubernetes, CI/CD pipelines, MLflow, Terraform, MLOps platforms, agentic AI frameworks such as CrewAI, AutoGen and LangGraph, knowledge graphs, and NLP and machine-learning fundamentals.

Preferred skills: Basic artificial intelligence, machine learning and Generative AI for data analytics.

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

Skills for this role

Generative AILarge Language ModelsLLMOpsMLOpsRetrieval-Augmented GenerationDockerKubernetesCI/CDMLflowTerraformCrewAIAutoGenLangGraphKnowledge GraphsNatural Language ProcessingMachine LearningModel MonitoringModel EvaluationResponsible AIAPI IntegrationCloud ServicesCommunicationCollaborationProblem-solving

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

Deliver LLM applications through production
The role designs, develops, deploys and maintains GenAI applications, with LLM pipelines spanning development, testing and production. Prepare to explain end-to-end engineering decisions, not just a prototype.
Make enterprise knowledge usable through RAG
The posting calls for RAG solutions using enterprise knowledge repositories and experience with assistants, chatbots, search or knowledge-management solutions. Prepare a concrete account of how retrieval and generated answers would be evaluated.
Integrate with existing systems
LLMs must connect with APIs, databases, business applications and cloud services. Prepare to describe integration boundaries, failure handling and how an application would fit into an existing workflow.
Operate and govern the LLM lifecycle
LLMOps duties include deployment, monitoring, evaluation, versioning and CI/CD. The role also monitors performance, latency, cost, hallucinations and drift. Prepare to explain what you would measure and how findings would inform changes.
Coordinate secure, scalable delivery
The engineer collaborates with data scientists, ML engineers and cloud teams while ensuring security, compliance, observability and responsible AI practices. Prepare to explain how technical decisions and operational responsibilities would be shared across those teams.

Likely challenges to think through

Balancing answer quality with operating cost
A likely trade-off, not a confirmed Infosys problem: improving retrieval or generation quality may add latency and cost, all of which the posting says to monitor. Prepare an example of comparing quality, latency and cost before choosing a change.
Using enterprise knowledge without overexposing it
A likely difficulty is making repository content useful for RAG while meeting the stated security and compliance responsibilities. Prepare to discuss access controls, data boundaries and tests for inappropriate disclosure as design considerations, not assumed employer procedures.
Changing models without losing control
Frequent iteration can conflict with reproducible evaluation and stable production behavior. Because the role covers CI/CD, versioning, drift monitoring and lifecycle management, prepare to describe how you would test, track and, if needed, reverse an LLM application change.

Practice your answers

Technical practice

How would you take an LLM application from development through testing to production and ongoing maintenance?

What to coverThe posting calls for LLM applications and pipelines across all three environments. Suggested preparation: explain how you would validate changes, manage releases and maintain the application after deployment.

How would you design a RAG solution over an enterprise knowledge repository?

What to coverRAG using enterprise knowledge repositories is an explicit responsibility. Suggested preparation: discuss document preparation, retrieval relevance, grounding answers in retrieved material and how you would test failures.

How would you integrate an LLM application with APIs, databases and business applications?

What to coverThe role explicitly includes these integrations and cloud services. Suggested preparation: describe data flow, interface failures and how you would limit access to information the application is permitted to use.

How would you evaluate answer quality and detect hallucinations in an enterprise AI assistant?

What to coverThe posting calls for model evaluation, hallucination monitoring and experience with assistants or similar applications. Suggested preparation: explain representative test cases, human review where useful and the trade-off between answer quality and coverage.

A production LLM application becomes slower and more expensive; how would you investigate?

What to coverLatency, performance and cost optimization are named monitoring concerns. Suggested preparation: identify measurements you would compare, isolate the affected pipeline stage and assess the quality impact of any optimization.

How would you automate and version LLM deployments using CI/CD and MLflow?

What to coverThe posting specifies automated CI/CD for model deployment and versioning and lists MLflow. Suggested preparation: describe what you would track, test and promote between environments, including how you would recover from a poor release.

What roles would Docker, Kubernetes and Terraform play in deploying a scalable LLM application?

What to coverAll three technologies are listed, and the role calls for scalable AI solutions. Suggested preparation: distinguish packaging, orchestration and infrastructure provisioning, then explain operational trade-offs without assuming a particular cloud provider.

How would you address security, compliance, observability and responsible AI in an LLM platform?

What to coverThe posting explicitly requires attention to all four areas. Suggested preparation: discuss access controls, handling of sensitive data, useful operational signals and checks for harmful or unsupported outputs.

When would an agentic AI framework be useful, and when would a simpler LLM pipeline suffice?

What to coverCrewAI, AutoGen and LangGraph appear in the technical requirements. Suggested preparation: compare multi-step coordination with a simpler workflow, considering reliability, observability, latency and cost; do not assume the role mandates a specific framework.

How might knowledge graphs and NLP fundamentals inform enterprise search or retrieval?

What to coverThe posting lists knowledge graphs and NLP and machine-learning fundamentals alongside RAG and search solutions. Suggested preparation: explain when explicit relationships or language-processing methods could improve retrieval, and how you would verify that they help.

Behavioural practice

Tell me about a time you coordinated with specialists from different technical teams to deliver a solution.

What to coverThe role calls for collaboration with data scientists, ML engineers and cloud teams. Suggested preparation: use a real situation, your actions to resolve dependencies or differing priorities, and the result.

Describe a time you diagnosed and resolved a production quality or performance problem.

What to coverMaintaining applications and monitoring performance, latency and hallucinations are stated responsibilities. Suggested preparation: give the situation, the evidence you gathered and actions you took, then report the outcome without claiming an improvement you cannot substantiate.

Tell me about a time you had to explain an AI system's limitations or trade-offs to others.

What to coverThe posting requests effective communication and analytical skills and names quality, cost and responsible AI concerns. Suggested preparation: describe the audience and situation, how you communicated the evidence and options, and what resulted.

Describe a time you worked through a security or governance concern during delivery.

What to coverSecurity, compliance, governance and responsible AI are explicit aspects of the role. Suggested preparation: recount a real situation, the actions you took with relevant collaborators, and the result; avoid implying that a particular company process applies here.

Sharpen your knowledge

Topics to prepare

LLM lifecycle and production deployment
Suggested exercise: Diagram how an LLM application moves through development, testing, production, versioning and rollback; explain what you would check at each stage. Relevance: The posting explicitly calls for LLM lifecycle management and production deployment.
RAG and enterprise knowledge
Suggested exercise: Sketch a retrieval-and-answer flow for a knowledge repository, then explain how you would test whether an answer is supported by retrieved material. Relevance: The role explicitly develops RAG solutions using enterprise knowledge repositories and lists knowledge graphs as a technical requirement.
Evaluation and operational trade-offs
Suggested exercise: Compare two hypothetical LLM releases on answer quality, hallucinations, latency and cost; decide which you would deploy and why. Relevance: The posting calls for evaluation and monitoring of performance, latency, costs, hallucinations and model drift.
Security, governance and responsible AI
Suggested exercise: Review a hypothetical assistant that accesses internal documents; identify access, disclosure and audit risks and propose controls. Relevance: The role explicitly requires security, compliance, observability, governance and responsible AI practices.
Agentic AI design
Suggested exercise: Explain when a multi-step agent is preferable to a simpler LLM workflow, and where you would limit its actions. Relevance: CrewAI, AutoGen and LangGraph are listed technical requirements; this design comparison is preparation, not a stated employer task.
NLP and machine-learning fundamentals
Suggested exercise: Explain how retrieval relevance, input ambiguity and changing data can affect an assistant’s output, distinguishing these from model drift. Relevance: NLP and machine-learning fundamentals are explicitly listed.

Skills to revise

Build and test a small RAG assistant
Suggested exercise: Use a sample document collection to implement retrieval and LLM-based answering, then test questions with and without supporting documents. Relevance: The posting calls for RAG and experience with enterprise assistants, search or knowledge-management solutions; the sample is only a practice exercise.
Integrate an LLM application with data and services
Suggested exercise: Connect a practice assistant to an API and a database, and test unavailable-service and empty-result cases. Relevance: The role explicitly integrates LLMs with APIs, databases, business applications and cloud services.
Package and provision a deployment
Suggested exercise: Package a practice LLM service with Docker, outline its Kubernetes deployment, and write a small Terraform configuration for supporting infrastructure. Relevance: Docker, Kubernetes and Terraform are listed technical requirements.
Automate deployment and version tracking
Suggested exercise: Draft a CI/CD pipeline that tests and deploys a versioned application, and record a practice model run in MLflow. Relevance: The posting calls for automated deployment and versioning and lists CI/CD, MLflow and MLOps platforms.
Evaluate and monitor an application
Suggested exercise: Run a fixed question set against two application versions; record answer quality, hallucination cases, latency and cost, then state what changes would trigger investigation. Relevance: Evaluation and monitoring of these measures are explicit responsibilities.
Try a listed agentic framework
Suggested exercise: Implement a bounded two-step practice workflow in one of CrewAI, AutoGen or LangGraph and test a failed step. Relevance: These frameworks are explicitly listed; the particular workflow is a suggested exercise, not a stated project.

Your preparation checklist

Map evidence to the posting
Prepare a one-page evidence sheet for the listed education, experience ranges, production LLM work, and technical tools. The posting states 3–5 years of work experience and separately 2–5 years in AI/ML, GenAI, LLM engineering, or MLOps; distinguish direct experience from familiarity.
Build a project walkthrough
Prepare a concise walkthrough of an actual enterprise assistant, chatbot, search, or knowledge-management project if available. Show the problem, your contribution, architecture, deployment, results, and trade-offs; label any hypothetical example clearly.
Practice a RAG design exercise
Sketch how an LLM application could retrieve from an enterprise knowledge repository and connect to an API or database. Explain retrieval quality, access controls, failure handling, and when a knowledge graph or listed agentic framework would add value.
Outline a production deployment
Draw a development-to-testing-to-production flow with model and application versioning, CI/CD, and rollback. Be ready to explain where Docker, Kubernetes, MLflow, Terraform, and an MLOps platform could fit without assuming Infosys uses a particular setup.
Prepare an evaluation and monitoring plan
Create a small example test set and a monitoring table covering answer quality, hallucinations, latency, cost, performance, and drift. Explain how findings would trigger investigation or a release decision.
Rehearse governance and collaboration trade-offs
Prepare one example or clearly labeled scenario showing security, compliance, observability, and responsible-AI decisions, plus how you would coordinate with data scientists, ML engineers, and cloud teams.
Write questions for the employer
Ask which enterprise knowledge sources and integrations matter most, how GenAI applications are evaluated and governed, and what deployment and monitoring responsibilities this engineer would own.

Availability last checked: 4 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…