New opportunity

Applied AI Engineer, Vikasit Sarkar

LambdaQ Labs · Pune / Delhi, India

About this role

Build Vikasit Sarkar, an AI system that helps Indian government officials draft tenders, policies and letters in 22 Indian languages using their departments’ knowledge. The role combines applied AI, data engineering and deployment with government departments. It is a full-time position in Pune or Delhi, India.

Responsibilities

Build knowledge-grounded drafting systems for government workflows.
Work with departments to capture rules, precedents and document formats.
Deliver multilingual experiences, including Marathi, Hindi and English, that officials can trust.
Own pilots from data preparation through deployment and feedback.

Requirements

Production experience in applied machine learning or LLM application engineering.
Strong Python and data-pipeline skills.
Ability to work with non-technical stakeholders and messy real-world data.

Preferred

Indian-language NLP; GovTech or regulated-industry experience; and Marathi or Hindi fluency. LambdaQ Labs states that it hires for skill over credentials and welcomes links and projects in applications.

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

Skills for this role

Applied Machine LearningLLM Application EngineeringPythonData PipelinesData EngineeringMultilingual NLPIndian-language NLPStakeholder ManagementGovTechMarathiHindi

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

Ground drafting in departmental knowledge
The role builds systems that help officials draft tenders, policies and letters using their departments’ knowledge. Prepare to explain how a production LLM application can use relevant source material and make draft claims reviewable.
Capture rules and document conventions
The engineer works with departments to capture rules, precedents and document formats. Prepare an approach for eliciting these from non-technical stakeholders and turning messy documents into usable data.
Deliver trustworthy multilingual experiences
The posting calls for drafting across 22 Indian languages, including Marathi, Hindi and English. Prepare to discuss how to check meaning and document conventions across languages; Indian-language NLP experience and Marathi or Hindi fluency are preferred, not required.
Own pilots through deployment and feedback
The role spans data preparation, applied AI development, deployment and feedback. Prepare a concise account of how you would structure an end-to-end pilot using Python and data pipelines, both stated requirements.

Likely challenges to think through

Grounding versus incomplete source material
A likely challenge is producing useful drafts when departmental rules or precedents are hard to extract, without presenting unsupported content as authoritative. Prepare to discuss source traceability, missing-information handling and human review as proposed safeguards, not stated company processes.
Language coverage versus drafting quality
Supporting 22 languages may make it difficult to preserve the meaning and format of official documents across languages. Prepare a proposed evaluation approach using language-specific examples and reviewer feedback, while distinguishing broad coverage from proven quality.
Pilot speed versus dependable deployment
End-to-end pilot ownership may require balancing quick iteration with reliable data pipelines and feedback from officials. Prepare to explain how you would choose a narrow first workflow, test it with stakeholders and use findings to prioritize changes; these are preparation suggestions, not a described pilot plan.

Practice your answers

Technical practice

How would you design a drafting system grounded in a department’s own knowledge?

What to coverPreparation suggestion: Explain how relevant departmental material could be selected and used to support a draft, and how an official could check which material informed it. Discuss the trade-off between producing a useful draft and avoiding unsupported statements.

How would you handle conflicting rules and precedents?

What to coverPreparation suggestion: Describe how you would identify conflicts, establish which source applies, and avoid silently choosing an uncertain interpretation. The role calls for capturing departmental rules and precedents; do not assume a particular approval process.

How would you prepare messy departmental documents for an AI application?

What to coverPreparation suggestion: Walk through a Python data pipeline for extracting, cleaning, organizing, and checking documents while preserving information needed for grounding. Explain how you would detect errors that could affect draft accuracy.

How would you support tenders, policies, and letters with different formats?

What to coverPreparation suggestion: Discuss how to capture each document type’s structure and departmental conventions without forcing them into one rigid template. Explain how you would test that the output follows the relevant format and stays grounded.

How would you assess drafts across Marathi, Hindi, English, and other Indian languages?

What to coverPreparation suggestion: Address meaning, terminology, formatting, and fidelity to departmental sources, rather than fluency alone. Explain how you would investigate a draft that reads naturally but changes the meaning of a rule.

How would you evaluate whether officials can trust the drafting experience?

What to coverPreparation suggestion: Propose checks for factual support, adherence to rules and formats, and usefulness to the official reviewing the draft. Discuss how feedback from officials could expose failures that automated checks miss.

How would you troubleshoot an incorrect draft in a deployed pilot?

What to coverPreparation suggestion: Trace whether the problem arose in document preparation, selection of departmental knowledge, or draft generation. Explain how you would reproduce the issue, correct it, and check for similar errors.

What trade-offs would you make when taking a pilot from data preparation to deployment?

What to coverPreparation suggestion: Describe how you would decide what to validate before officials use the system and what to monitor through feedback afterward. Keep the answer tied to the role’s stated end-to-end pilot ownership, without assuming a specific deployment setup.

Behavioural practice

Tell me about a time you translated non-technical stakeholders’ needs into system requirements.

What to coverSuggested practice: Use a real situation, the questions you asked and actions you took to resolve ambiguity, and the resulting change in the work. Connect your example to capturing rules, precedents, or document formats without claiming experience you do not have.

Describe a time messy data threatened a project’s outcome.

What to coverSuggested practice: Explain the situation, how you identified and prioritized the data problems, what you did, and the result. Include how you communicated any limits the remaining data imposed.

Tell me about a project you owned from preparation through delivery and feedback.

What to coverSuggested practice: Set out your responsibilities, the decisions you made across stages, and the outcome. Focus on coordination and course corrections relevant to the role’s end-to-end pilot ownership.

Describe a time user feedback challenged something you had shipped.

What to coverSuggested practice: Give the situation, how you investigated and responded to the feedback, and what changed as a result. For this role, a useful example would show how you balanced user trust with evidence about the system’s behavior.

Sharpen your knowledge

Topics to prepare

Knowledge-grounded government drafting
The role builds drafts of tenders, policies and letters using departmental knowledge. Revise how rules, precedents and document formats should constrain a draft; as practice, identify which statements in a sample letter need supporting departmental material.
Production applied AI
Production applied ML or LLM application engineering is required. Revisit how to assess output quality and handle unsupported or inconsistent drafts; as practice, define checks for a hypothetical policy-drafting workflow.
Multilingual drafting
The posting calls for experiences across 22 Indian languages, including Marathi, Hindi and English; Indian-language NLP is preferred. Compare a short draft across those three languages for meaning, terminology and document-format consistency.
End-to-end pilots
The engineer owns pilots from data preparation through deployment and feedback. Sketch the stages of a hypothetical departmental pilot and what evidence at each stage would show whether officials can use the drafts.

Skills to revise

Python data preparation
Strong Python and data-pipeline skills are required. Practice turning a small, messy collection of sample rules and letters into consistently structured records, while flagging missing fields and duplicates.
Grounded drafting implementation
The role builds knowledge-grounded drafting systems. As a practice exercise, implement a small draft-generation flow using sample departmental material and inspect whether each substantive claim is supported by that material.
Departmental requirements elicitation
Working with departments and non-technical stakeholders is explicit. Rehearse questions that uncover applicable rules, useful precedents, required letter formats and how an official would judge a draft.
Pilot feedback triage
The posting assigns ownership through deployment and feedback. Practice sorting hypothetical official feedback into data, drafting and language issues, then propose a specific next change for each.

Your preparation checklist

Prepare a production AI project walkthrough
The role asks for production applied ML or LLM application engineering. Prepare a concise walkthrough of a relevant system you built, your decisions, deployment responsibilities, and what you measured. If you have no directly relevant project, use a clearly labeled design example instead.
Sketch a knowledge-grounded drafting workflow
For a tender, policy, or letter, draw a proposed flow from departmental documents to a reviewable draft. Explain how you would surface supporting rules and precedents and handle missing or conflicting information; present these as design choices, not LambdaQ Labs’ existing process.
Show a Python data-pipeline example
The posting calls for strong Python and data-pipeline skills with messy data. Prepare a code sample or walkthrough showing how you parsed, cleaned, validated, and updated documents while preserving their source and format.
Plan a multilingual evaluation exercise
The role covers 22 Indian languages and specifically mentions Marathi, Hindi, and English. Outline sample checks for meaning, terminology, document format, and unsupported claims across languages. Distinguish any languages you have worked with from those you would need to test with reviewers.
Map a department pilot from discovery to feedback
Draft a one-page proposed pilot plan: questions to elicit rules and document formats from non-technical officials, a small data-preparation scope, deployment considerations, and a way to collect feedback. Label it as your approach; the posting does not prescribe a pilot method.
Prepare evidence and employer questions
Select relevant project links or artifacts, which the posting welcomes, and note what each demonstrates. Ask how departments provide knowledge, how officials review multilingual drafts, and what would define a successful pilot.

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…