Envisstack builds AI systems that work inside Indian businesses: from the first workshop to production, and the support after. We are a forward-deployed AI engineering firm in Mumbai. We sit with the client, learn the real workflow, connect AI to the systems they already run (WhatsApp, Tally, SAP, Salesforce, Zoho, core banking, legacy databases), prove results against an agreed metric, and keep the system working.

The role

You build the AI systems we deliver to clients and keep them working after go-live. You own components end to end: retrieval pipelines, document extraction, agent tools, integrations with client systems, and the evaluations that prove they work. You work directly with the founders.

We welcome career switchers

Many of the best AI engineers in India today started as Java, .NET, Python or QA engineers in IT services or product companies. If you have two to five years of software engineering and have been building with LLMs on the side, you are who we want. You do not need an ML degree.

What you will do

- Build retrieval-augmented generation (RAG) systems, extraction pipelines and tool-using agents on real client data.
- Integrate with client systems: WhatsApp Business API, Tally, Zoho, SAP, Salesforce, Google Workspace, Microsoft 365, and their databases.
- Design and run evaluations with the client: success metrics, evaluation sets, scorers, thresholds.
- Add guardrails, logging, cost limits and fallbacks to everything you ship.
- Diagnose production failures from traces and turn them into new test cases.
- Join discovery workshops and explain technical options to client teams in plain language.

What we look for

- Two or more years of skilled software engineering, or strong project evidence.
- Strong Python. Comfortable with APIs, SQL, Git, Docker and one cloud (AWS, Azure or Google Cloud).
- Hands-on experience with LLM APIs: prompting, structured output, tool use, embeddings, vector search.




- You test your own code and review AI-generated code with care.
- You think about evaluation before you think about frameworks.
- Clear written English. An Indian language is a plus, especially for voice and Indic work.

Nice to have

- Experience with Indian enterprise systems (Tally, SAP, core banking, hospital information systems).
- Experience with Langfuse, Promptfoo or another evaluation or observability tool.
- Awareness of the DPDP Act and sector rules (RBI, IRDAI, SEBI).

Interview process

1. 20-minute call.
2. Take-home (4 to 6 hours, time-boxed): build a small RAG system and write an evaluation report.
3. AI-assisted coding round. Use your own coding agent. We watch how you review and test its output.
4. Eval-design exercise.
5. Review of your take-home, plus a deep-dive on past work.
6. Values conversation.

An honest note

Envisstack is a new company. We do not have a client list or a team yet, and you would be one of the first people here. That means real ownership, direct work with the founders, and the chance to shape how we work.

It also means less structure and more uncertainty than an established firm. The legal entity, Envisstack India Private Limited, is being formed, and this role starts once it is incorporated and our first client engagement is signed. We will tell you the expected start date and the full terms before any offer.

We would rather you know all of this now.

How to apply

Apply here, or email [email protected] with the role title in the subject line. Send a CV or LinkedIn profile, and a short note (five to ten lines) about something you built or fixed that you are proud of. A link to code, a write-up or a demo helps but is not required. We reply to every application within ten working days.

We hire on ability. We welcome applicants of every gender, caste, religion, region, disability status, sexual orientation and background. If you need any adjustment during the interview process, tell us and we will make it.

📌 Artificial Intelligence Engineer (Mumbai)
🏢 Envisstack
📍 Mumbai