New opportunity

Senior AI Engineer

EMB Global · Gurgaon, India

About this role

EMB Global seeks a Senior AI Engineer in Gurgaon, Haryana, India. The posting lists 6-6 years of experience and identifies the role as full-time. The engineer will design and deliver production-ready AI/ML solutions for enterprise clients, translating business problems into technical architectures and working with stakeholders to identify measurable business value.

Responsibilities

Define solution architectures, technology stacks, APIs, data flows and integration patterns. Evaluate LLMs, AI models, frameworks and cloud technologies for enterprise use cases.
Lead development of generative AI applications using LLMs, retrieval-augmented generation (RAG), AI agents, embeddings, vector databases and prompt engineering. Build and optimise agents and multi-agent workflows, working with models from OpenAI, Anthropic, Google and open-source ecosystems.
Own projects from proof of concept and prototyping through production deployment. Develop maintainable code, AI pipelines, model-serving layers, APIs, integrations and backend components; establish testing and validation frameworks.
Measure and benchmark model quality, accuracy, relevance, hallucination, latency and cost. Improve prompts, RAG pipelines, model selection, inference and workflows for reliable, cost-efficient production use.
Design and deploy AI workloads on AWS, Azure or GCP. Work with containerization, databases, CI/CD, monitoring and scalable infrastructure alongside DevOps and cloud teams, meeting enterprise performance, security, scalability and availability needs.
Apply security, privacy, governance and compliance practices. Identify risks involving hallucination, bias, prompt injection and unauthorised access, and implement safeguards and monitoring.

Leadership and collaboration:

Mentor AI engineers and developers; lead code and architecture reviews, technical design discussions, engineering standards and documentation. Break complex requirements into deliverable tasks.
Work with clients, business stakeholders and internal teams in solution workshops, presentations and demonstrations, explaining technical approaches in business terms. Assess emerging AI tools and models, build rapid proofs of concept, and contribute reusable frameworks and solution assets.

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 ModelsMachine LearningAI AgentsMulti-agent workflowsRetrieval-Augmented GenerationEmbeddingsVector databasesPrompt engineeringAI solution architectureModel evaluationModel servingAPI developmentCloud deploymentAWSAzureGCPContainerizationCI/CDAI governanceData privacyPrompt injection mitigationTechnical leadershipStakeholder communication

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

Client-led AI solution architecture
Translate enterprise business problems into practical AI solutions with measurable value; define the architecture, technology stack, APIs, data flows and integration patterns.
Generative AI application delivery
Lead development of LLM applications using RAG, embeddings, vector databases and prompt engineering, including agents and multi-agent workflows.
Ownership through production
Take projects from proof of concept to deployment, building maintainable pipelines, model-serving layers, APIs and backend integrations. Work with cloud and DevOps teams on scalable infrastructure and operations.
Evaluation and responsible AI
Establish testing and benchmarks for quality, relevance, hallucination, latency and cost. Apply security, privacy and governance safeguards, including attention to bias, prompt injection and unauthorised access.
Technical leadership
Mentor engineers, lead code and architecture reviews, set engineering practices and documentation, and turn complex requirements into deliverable tasks.

Likely challenges to think through

Balancing model quality with operating cost
Preparation hypothesis: Better answers may increase latency or cost, while cheaper models may reduce relevance or reliability. Prepare a concrete approach to benchmarking models, prompts and RAG changes against these measures.
Moving quickly without weakening production controls
Preparation hypothesis: Rapid proofs of concept and emerging tools may be difficult to reconcile with maintainability, testing, monitoring and security. Prepare to explain how you would decide what must change before an experiment is deployed.
Making agent workflows safe for enterprise data
Preparation hypothesis: More capable agents and integrations can expand the risk of prompt injection, data exposure or unauthorised actions. Prepare to discuss safeguards, access boundaries and monitoring while preserving useful functionality.

Practice your answers

Technical practice

How would you turn an enterprise client's business problem into a production-ready AI architecture?

What to coverSuggested answer focus: Start with the business outcome and a measurable success criterion, then explain how you would define data flows, APIs, integrations and the technology stack. Discuss trade-offs among reliability, scalability, security, latency and cost.

How would you design and evaluate a retrieval-augmented generation pipeline?

What to coverSuggested answer focus: Explain the roles of embeddings, a vector database, retrieval and prompting. Describe how you would test whether retrieved information is relevant and whether the final answer is accurate and grounded.

When would you use an AI agent or a multi-agent workflow instead of a simpler application?

What to coverSuggested answer focus: Compare the value of delegated steps or tool use with added complexity, latency and failure modes. Explain how you would constrain, test and monitor the workflow for an enterprise use case.

How would you select an LLM or framework for an enterprise use case?

What to coverSuggested answer focus: Describe a comparison of suitable provider and open-source options against the use case's quality, privacy, latency, cost and integration needs. Explain why a small, representative proof of concept would be more useful than a choice based on a single benchmark.

What evaluation framework would you establish before deploying a GenAI application?

What to coverSuggested answer focus: Define representative test cases and measures for quality, accuracy, relevance, hallucination, latency and cost, as listed in the posting. Explain how you would compare versions and use failures to guide improvements.

How would you investigate a rise in hallucinations in a deployed RAG application?

What to coverSuggested answer focus: Separate retrieval failures from prompt, model and workflow failures using examples and evaluation results. Discuss targeted fixes and how you would verify that they improve answers without unacceptable latency or cost.

How would you build maintainable APIs and model-serving components for an AI application?

What to coverSuggested answer focus: Describe clear boundaries among pipelines, serving, APIs and integrations, with testing and validation appropriate to each. Explain how those choices support changes from prototype through production.

How would you deploy and operate an AI workload in a cloud environment?

What to coverSuggested answer focus: Discuss how containerization, CI/CD, databases and monitoring fit a deployment on AWS, Azure or GCP. Address collaboration with DevOps and cloud teams and trade-offs involving performance, availability and scalability.

What safeguards would you design for an enterprise AI application?

What to coverSuggested answer focus: Consider data privacy, unauthorised access, prompt injection, bias and hallucination—the risks named in the posting. Explain where safeguards and monitoring belong in the data flow and how you would check that they work.

How would you decide whether a GenAI proof of concept is ready for production?

What to coverSuggested answer focus: Compare the proof of concept with measurable business value and evidence from testing, validation and operational monitoring. Identify what must change in code, integrations and safeguards before deployment, and what might be reusable in a future solution.

Behavioural practice

Tell me about a time you explained a complex AI approach to a business stakeholder.

What to coverSuggested answer focus: Give the situation, the stakeholder's decision or concern, the action you took to explain trade-offs in business terms, and the result. Use a real example rather than assuming a particular client or outcome.

Describe a time you mentored engineers or improved a technical design through review.

What to coverSuggested answer focus: Set out the team's situation, your actions in a code or architecture review, and the result for delivery or engineering quality. Make your own contribution and the team's contribution clear.

Tell me about a time you coordinated with another team to move a solution into production.

What to coverSuggested answer focus: Describe the situation, what you did to resolve deployment or operational dependencies, and the result. A real example involving cloud or DevOps collaboration would connect directly to the advertised work.

Describe a time an AI proof of concept did not initially meet its success criteria.

What to coverSuggested answer focus: Explain the original goal, the evidence that exposed the gap, your actions and the result. Discuss how you balanced model quality with practical constraints such as latency or cost, if those were part of your example.

Tell me about a time you evaluated an emerging AI tool or model for a practical use case.

What to coverSuggested answer focus: Describe the situation, how you tested its usefulness, what you recommended and the result. If applicable, explain whether the work produced a reusable approach rather than implying that adoption was inevitable.

Sharpen your knowledge

Topics to prepare

Enterprise AI solution architecture
Review how business objectives translate into architecture, APIs, data flows and integrations. Exercise: sketch two architecture options for an enterprise AI use case and explain the trade-offs against a measurable business outcome. The role explicitly owns solution design and technology selection.
RAG and agent design
Review when to use retrieval, embeddings, vector databases, a single agent or a multi-agent workflow. Exercise: diagram how each option would handle the same information-retrieval task and identify likely failure points. All are named in the GenAI responsibilities.
Model evaluation and optimisation
Review how to assess quality, relevance, hallucination, latency and cost together. Exercise: define a small comparison rubric for two model-and-prompt options, including what would trigger a change. The posting calls for benchmarking and production reliability.
Production and cloud trade-offs
Review the path from proof of concept to deployment, including model serving, scalability, availability and monitoring. Exercise: outline deployment decisions for one AI workload on AWS, Azure or GCP and explain the operational trade-offs. These platforms and concerns are explicitly listed; no particular platform is mandated.
AI governance and security
Review privacy, bias, prompt injection and unauthorised-access risks. Exercise: make a threat-to-safeguard table for an enterprise RAG application, with a monitoring signal for each risk. The role explicitly calls for safeguards and monitoring.
Technical leadership and client value
Review how to turn ambiguous client needs into deliverable tasks and explain technical choices in business terms. Exercise: draft a brief solution-workshop agenda, success measure and architecture-review questions. The posting includes client workshops, mentoring and design reviews.

Skills to revise

Build a small RAG workflow
As a revision exercise, assemble a document-to-embedding-to-vector-search-to-LLM flow, then test responses when evidence is missing. This practises the RAG components the role explicitly calls for; the exercise is a preparation suggestion, not a specified employer project.
Implement an AI-serving API
Practise wrapping a model workflow in an API with input validation, error handling and tests; document its data flow and integration boundary. The posting calls for maintainable code, model-serving layers, APIs and validation frameworks.
Run a comparative evaluation
Create a small test set and record relevance, unsupported claims, latency and cost for two prompt or model choices; explain the resulting trade-off. This directly rehearses the stated benchmarking and optimisation work.
Prepare a production deployment plan
Practise containerising a small AI service and outlining CI/CD checks, monitoring and a rollback path for deployment on one listed cloud platform. Containerization, CI/CD, cloud deployment and monitoring appear in the posting; this particular exercise is a suggestion.
Review an agent workflow for safety
Test a sample agent workflow with prompt-injection attempts and requests for unauthorised information; revise its safeguards and record what should be monitored. The role calls for agent optimisation and mitigation of these enterprise risks.

Your preparation checklist

Prepare an end-to-end project walkthrough
The role covers work from proof of concept through production. Prepare one example you can substantiate, showing the business problem, your decisions, APIs and integrations, testing, deployment, and measurable results. If you lack a comparable project, label the walkthrough as a proposed approach.
Sketch an enterprise GenAI architecture
For a hypothetical client use case, draw data flows, APIs, a RAG pipeline and any justified agent workflow. Be ready to explain model, embedding and vector-database choices, including when a simpler design would suffice.
Build an evaluation plan
The posting calls for benchmarking quality, relevance, hallucination, latency and cost. Prepare a small test-set design, metrics, comparison table and decision rule for improving prompts, retrieval or model selection.
Map the path to production
Outline how you would package and serve the solution on AWS, Azure or GCP, with CI/CD, monitoring and capacity considerations. Add a risk register covering privacy, prompt injection, bias and unauthorised access, with proposed safeguards.
Gather leadership evidence
Prepare concise, factual examples of technical direction, code or architecture reviews, mentorship, and breaking requirements into deliverable tasks. Note the decisions you made and their observable outcomes.
Rehearse a client-facing explanation
The role includes workshops, presentations and demonstrations. Prepare a short explanation that connects a proposed AI solution to a business measure, states its limitations and invites stakeholder input.
Draft questions for EMB Global
Ask how this role balances client discovery, prototyping and production ownership; how AI quality and business value are measured; and how engineering, cloud and client teams share deployment and governance decisions.

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…