AI Engineer Mumbai, hybrid, client sites 2 to 5 years

Envisstack; Mumbai; Hybrid; INR 1,500,000 - 2,300,000; Full time

Job summary

Envisstack in Mumbai seeks an AI engineer to design and build end-to-end AI features for client engagements, from pilot projects to production and managed retainers. You will work on WhatsApp-based support agents, document pipelines that read claims and invoices, and knowledge assistants over SOPs, with hybrid work and time on client sites.

The role emphasizes data, integration, edge cases and measurement. You will design prompts and context, implement retrieval, tool calling, and agent

Qualifications

• 2 to 5 years of software engineering experience, with at least one system you built running in production.
• Strong Python, and comfort with APIs, SQL, Git, testing and Docker.
• Hands-on work with model APIs: prompting, structured outputs, tool calling, and handling failures and rate limits.
• Experience building or improving a retrieval system: parsing, chunking, embeddings, vector search and reranking.
• A habit of measuring. You can describe how you knew a change was an improvement.
• Clear communication in English with technical and non-technical people.

Responsibilities

• Design and build AI features end to end: prompts and context, retrieval, tool calling, agent workflows, and the services around them.
• Build and own evaluation sets with client experts. Run evaluations on every change and report the scores honestly.
• Integrate with client systems through APIs, databases and files: WhatsApp Business Platform, CRMs, Tally, SAP, core banking and document stores.
• Compare models, commercial and open-weight, on accuracy, latency and cost, and recommend one with evidence.
• Add guardrails, personal-data redaction, logging and monitoring before anything reaches users.
• Deploy to client cloud accounts and on-premise environments, working with their IT and security teams.
• Spend time on site with users. Watch them use the system, collect failures, and fix them.
• Write clear design notes, runbooks and handover documents.

Skills

Python

APIs

SQL

Git

Testing

Docker

Model APIs

Retriever

Tools

Git

Docker

APIs

Job description

Location Mumbai, hybrid, with time at client sites

You will build the AI systems we deliver to clients: support agents on WhatsApp, document pipelines that read claims and invoices, knowledge assistants over SOPs, and agents that reconcile ledgers. You will work on one or two engagements at a time, from the pilot to production and into the managed retainer.

The model is rarely the hard part. The hard parts are the data, the integration, the edge cases and the measurement. This role is for an engineer who enjoys those.

• Design and build AI features end to end: prompts and context, retrieval, tool calling, agent workflows, and the services around them.
• Build and own evaluation sets with client experts. Run evaluations on every change and report the scores honestly.
• Integrate with client systems through APIs, databases and files: WhatsApp Business Platform, CRMs, Tally, SAP, core banking and document stores.
• Compare models, commercial and open-weight, on accuracy, latency and cost, and recommend one with evidence.
• Add guardrails, personal-data redaction, logging and monitoring before anything reaches users.
• Deploy to client cloud accounts and on-premise environments, working with their IT and security teams.
• Spend time on site with users. Watch them use the system, collect failures, and fix them.
• Write clear design notes, runbooks and handover documents.

You should have

• 2 to 5 years of software engineering experience, with at least one system you built running in production.
• Strong Python, and comfort with APIs, SQL, Git, testing and Docker.
• Hands-on work with model APIs: prompting, structured outputs, tool calling, and handling failures and rate limits.
• Experience building or improving a retrieval system: parsing, chunking, embeddings, vector search and reranking.
• A habit of measuring. You can describe how you knew a change was an improvement.
• Clear communication in English with technical and non-technical people.

Good to have

• Hindi, Marathi, Tamil or another Indian language, spoken and written.
• Experience with speech (ASR and TTS), document AI or OCR.
• Work with Indian enterprise systems: Tally, SAP, Finacle, Zoho, Salesforce or similar.
• Experience in a regulated sector such as banking, insurance or healthcare.
• Cloud experience on AWS, Azure or Google Cloud, ideally in Indian regions.
• Familiarity with agent frameworks and the Model Context Protocol (MCP).

First 30 days

Join a live engagement. Read the code and the evaluation set. Ship your first change to production with review.

By 60 days

Own a component, such as the retrieval layer or the integration with the client's CRM. Lead one evaluation review with the client.

By 90 days

Run a workstream on your own and present its weekly quality report to the client.