New opportunity

AI Developer

Wissen Technology Pvt. Ltd. · Mumbai, India

About this role

Wissen Technology is hiring an AI Developer to design, develop and deploy scalable AI-powered applications. The role focuses on production-grade generative AI systems using large language models (LLMs), retrieval-augmented generation (RAG) and agentic AI workflows. This is a full-time, hybrid position in Mumbai, India, requiring 6–10 years of experience.

Responsibilities

Build and deploy applications using LLMs such as GPT, Gemini, Llama, Mistral and Claude. Design RAG pipelines with vector databases and embeddings.
Develop AI agents and orchestrated workflows using LangChain, LangGraph, CrewAI, AutoGen or similar frameworks. Fine-tune, evaluate and optimize LLMs for production use.
Integrate AI services into scalable backend systems and APIs; work with structured and unstructured data for NLP and AI workflows; and deploy solutions on Azure, AWS or GCP.
Build observability, monitoring, guardrails and evaluation frameworks. Collaborate with product, engineering and business teams. For senior roles, mentor junior engineers and contribute to AI architecture decisions.

Required background:

Strong Python development skills, experience building production-grade GenAI systems, exposure to multi-agent workflows and AI observability tools, and an understanding of scalable backend architecture. The posting also calls for prior work on enterprise AI products or copilots.

Good to have:

Experience with complex multi-agent orchestration; LLM fine-tuning techniques including LoRA, PEFT and RLHF; and observability tools such as LangSmith, Weights & Biases, Arize and PromptLayer.
Knowledge of LLM evaluation, benchmarking, prompt testing, AI safety guardrails and hallucination reduction, plus data pipelines and ETL for AI/ML workloads.

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

Skills for this role

PythonGenerative AILarge Language ModelsRetrieval-Augmented GenerationAI agentsMulti-agent workflowsGPTGeminiLlamaMistralClaudeVector databasesEmbeddingsLangChainLangGraphCrewAIAutoGenLLM fine-tuningLLM evaluationLoRAPEFTRLHFNLPBackend architectureAPIsAzureAWSGCPAI observabilityLangSmith for AI observability and tracing tools may be long?

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-grade AI delivery
The role calls for Python-based GenAI applications integrated with scalable backends and APIs, then deployed on Azure, AWS or GCP. Prepare to explain architecture and deployment decisions from an application you built, rather than focusing only on a prototype.
RAG pipeline design
Designing retrieval-augmented generation with embeddings and vector databases is an explicit responsibility. Prepare to explain how you would handle structured and unstructured data and assess retrieval quality.
Agent and workflow orchestration
The posting calls for AI agents and orchestrated workflows using frameworks such as LangGraph, CrewAI or AutoGen, plus exposure to multi-agent workflows. Prepare to explain when orchestration adds value and how you would control workflow behavior.
Model quality and operational safeguards
The role includes fine-tuning, evaluating and optimizing LLMs for production, alongside monitoring, observability and guardrails. Prepare to describe measurable evaluation and what you would monitor after deployment; specific fine-tuning techniques are listed as good to have.
Cross-functional delivery
The developer collaborates with product, engineering and business teams on AI-driven solutions; prior enterprise AI product or copilot work is requested. Prepare to explain how technical choices connect to a business use case. Mentoring and architecture decisions apply to senior roles.

Likely challenges to think through

Retrieval quality versus system cost
A likely design trade-off is improving RAG answer quality without excessive retrieval latency or expense; this is a preparation hypothesis, not a stated employer problem. Prepare to compare chunking, retrieval and evaluation choices against those constraints.
Agent flexibility versus predictability
Agentic workflows may handle complex tasks but introduce more paths to test and monitor. As a preparation hypothesis, be ready to explain when a simpler workflow is preferable and how you would limit, trace and evaluate agent actions.
Rapid iteration versus production reliability
Changing prompts, models or fine-tuning approaches may improve results while making behavior harder to compare across releases. Prepare to discuss a repeatable evaluation, monitoring and guardrail approach; the posting requires these capabilities but does not describe a specific current issue.

Practice your answers

Technical practice

How would you choose an LLM for a production GenAI application?

What to coverThe role involves building applications with models such as GPT, Gemini, Llama, Mistral and Claude. Suggested preparation: explain how you would compare task quality, latency, cost, deployment constraints and evaluation results without assuming one model is best for every use case.

How would you design and troubleshoot a RAG pipeline?

What to coverThe posting calls for RAG using embeddings and vector databases. Suggested preparation: walk through ingestion, retrieval and answer generation, then explain how you would isolate retrieval failures from generation failures.

When would you use an agentic or multi-agent workflow?

What to coverThe role includes agent development and workflow orchestration, with complex multi-agent systems listed as good to have. Suggested preparation: describe task boundaries, state and handoffs, and when a simpler workflow would be easier to control than multiple agents.

How would you decide whether to fine-tune an LLM?

What to coverThe posting calls for fine-tuning and optimization; LoRA, PEFT and RLHF appear under good-to-have skills. Suggested preparation: compare fine-tuning with prompting or RAG, including data needs, evaluation effort and production maintenance.

How would you evaluate an LLM application before release?

What to coverThe role calls for evaluation frameworks, while benchmarking and prompt testing are listed as good to have. Suggested preparation: explain how you would build representative test cases, assess answer quality and compare changes against a baseline.

How would you integrate an AI service into a scalable Python backend?

What to coverStrong Python skills and integration with scalable backend systems and APIs are stated in the posting. Suggested preparation: discuss API boundaries, concurrency, timeouts, failures and how model latency affects the wider service.

How would you prepare structured and unstructured data for AI workflows?

What to coverThe role works with both data types, and data pipelines and ETL are listed as good to have. Suggested preparation: explain validation, transformation and data-quality checks, including how bad inputs could affect retrieval or model outputs.

What would you consider when deploying a GenAI service to the cloud?

What to coverThe posting names Azure, AWS and GCP as possible deployment platforms, without prescribing one. Suggested preparation: discuss scaling, access controls, reliability and operational cost at an architectural level.

How would you monitor and investigate a production GenAI issue?

What to coverObservability and monitoring are responsibilities, and the posting lists example AI observability tools as good to have. Suggested preparation: describe useful traces and metrics across retrieval, orchestration and model calls, and how you would locate a regression.

How would you reduce unsafe or unsupported model responses?

What to coverThe role includes guardrails; safety and hallucination reduction are listed as good to have. Suggested preparation: explain checks appropriate to the application, their limitations, and how you would test whether they improve outcomes without blocking useful responses.

Behavioural practice

Tell me about a time you aligned an AI solution with product and business needs.

What to coverThe posting calls for collaboration with product, engineering and business teams. Suggested preparation: use a real situation, your actions to clarify goals and trade-offs, and the resulting decision or outcome.

Tell me about a production AI issue you helped resolve.

What to coverThe role emphasizes production-grade systems, monitoring and evaluation. Suggested preparation: describe the situation, how you investigated and communicated it, what you did, and the result without overstating your individual contribution.

Tell me about a consequential AI architecture decision you contributed to.

What to coverContributing to architecture decisions is stated for senior roles. Suggested preparation: give a real example of the options considered, the evidence behind your recommendation, your role in the decision and its result.

Tell me about a time you mentored a junior engineer.

What to coverMentoring junior engineers is listed for senior roles. Suggested preparation: if applicable, describe the engineer's need, the specific support you provided and what changed as a result.

Tell me about a time data limitations changed your AI implementation plan.

What to coverThe role involves structured and unstructured data for NLP and AI workflows. Suggested preparation: if you have a relevant example, explain the data issue, how you worked with others to adapt the approach, and the resulting outcome.

Sharpen your knowledge

Topics to prepare

Production LLM choices
The posting calls for production-grade GenAI applications and names GPT, Gemini, Llama, Mistral and Claude as examples. Suggested exercise: compare two model options for an enterprise copilot, explaining how you would assess quality, latency and cost before deployment.
RAG design
The role explicitly includes RAG pipelines using embeddings and vector databases. Suggested exercise: diagram how unstructured documents become retrievable context, then explain how you would investigate an answer that cites irrelevant material.
Agent and multi-agent workflows
The posting requires exposure to multi-agent workflows and names LangChain, LangGraph, CrewAI and AutoGen as orchestration options. Suggested exercise: sketch an agent workflow with handoffs and failure paths, and explain when a simpler workflow would suffice.
Scalable AI architecture
AI services must integrate with scalable backend systems and APIs, with deployment on Azure, AWS or GCP. Suggested exercise: draw the components and data flow for a deployed AI API, including where requests, model calls and monitoring fit.
Evaluation, observability and guardrails
The responsibilities include evaluation, monitoring and guardrails; benchmarking, prompt testing and hallucination reduction appear under good-to-have skills. Suggested exercise: define a small test set and explain which failures and runtime signals you would track.
Fine-tuning decisions
Fine-tuning and optimizing LLMs are listed as responsibilities; familiarity with LoRA, PEFT and RLHF is good to have. Suggested exercise: explain when you would test retrieval or prompting before fine-tuning, and how you would compare the resulting model with a baseline.

Skills to revise

Python AI API implementation
Strong Python development and integration into backend APIs are explicit in the posting. Suggested exercise: implement a small Python API endpoint that accepts a request, calls an LLM and returns a structured response with clear error handling.
Build a retrieval pipeline
The role involves embeddings, vector databases and structured or unstructured data. Suggested exercise: ingest a small document set, retrieve relevant passages for test questions and inspect cases where retrieval misses the answer.
Implement an orchestrated workflow
Agent development is a stated responsibility. Suggested exercise: use one of the named frameworks to build a short, bounded workflow with a decision step and a recoverable failure, then trace its execution.
Run evaluations and inspect traces
The posting calls for evaluation frameworks and experience with AI observability tools; specific products such as LangSmith are examples under good-to-have skills. Suggested exercise: run a fixed set of prompts, record outputs and errors, and inspect traces to explain one failure.
Prepare a cloud deployment walkthrough
Deployment on Azure, AWS or GCP is a stated responsibility. Suggested exercise: prepare a walkthrough of how you would deploy and monitor the Python AI API on one of those platforms; choosing a particular service is preparation, not a stated requirement.

Your preparation checklist

Prepare a production GenAI project walkthrough
The posting asks for production-grade GenAI and prior enterprise AI product or copilot work. Prepare one relevant example showing the problem, your Python contributions, architecture, deployment decisions, results, and trade-offs; use only work you can substantiate.
Sketch a RAG design
The role includes embeddings, vector databases, and structured and unstructured data. Draw a proposed ingestion-to-answer pipeline and explain retrieval choices, source attribution, and how you would test answer quality.
Map an agent workflow
Agent orchestration and exposure to multi-agent workflows are listed. Prepare a workflow diagram showing agent responsibilities, tool access, state handling, failure paths, and when a simpler workflow would suffice.
Prepare an evaluation and operations plan
The role includes LLM evaluation, monitoring, observability, and guardrails. Draft a small test set and explain proposed quality and safety checks, trace review, and how findings would guide improvements.
Explain backend and cloud deployment choices
The posting calls for scalable APIs and deployment on Azure, AWS, or GCP. Sketch an API and deployment design covering data flow, access controls, latency, scaling, and operational trade-offs.
Collect evidence of technical and team decisions
Prepare concise examples of how you worked with product, engineering, or business teams and made architecture trade-offs. If applicable, include mentoring or model fine-tuning work; the posting associates mentoring with senior roles and lists techniques such as LoRA and PEFT as good to have.
Write questions for the employer
Ask which AI use cases are priorities, what data and deployment constraints apply, how production quality is measured, and what architecture or mentoring responsibilities this AI Developer role would carry.

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…