About the Product

You will be working on a multi tenant, AI first HR operations platform built for a US based company to run their back office. It is not a demo and not a wrapper. It handles real mailboxes, real cases, real client documents, and real implementation timelines that people are held to.

The core surfaces you would work on:

- Case management , driven by a delegation based Gmail mailbox model. Email lands, gets classified, gets linked to a client and a case, and produces a next best action.
- Implementation module , built on a canonical stage model of four arcs (Pre Stage 0 plus Stages 0 through 9). Every implementation is back scheduled from the client's first anticipated check date with configurable lead days per stage. The client needs to be able to add phases and checkpoints themselves as their internal process evolves.
- Client and document linkage , including provenance, dedup and merge logic, and visibility rules across tenants.
- Integrations: Gmail API with Pub/Sub push and OAuth token lifecycle, HubSpot, DocuSign, and LiveKit.

Stack: TypeScript, Next.js (App Router), Postgres via Supabase, Anthropic Claude API called directly, DigitalOcean, GitHub Actions.

What you will do

- Own product modules end to end. Postgres schema, migrations, API layer, and the Next.js dashboard on top of it. You design the tables, you write the queries, you build the UI, you ship the migration.
- Build AI powered workflows as engineering, not vibes. Prompt pipelines for triage, case intelligence, extraction, and next best action generation. Structured outputs with schema validation, evals and regression tests around prompt changes, and a explicit answer to "what happens when the model returns garbage".
- Be genuinely client facing. Join calls with US HR and ops stakeholders.



Learn their vocabulary, their compliance sensitivities, and their expectations on responsiveness. Reproduce reported issues, close the loop from complaint to deployed fix, and be able to explain the fix in plain English on the next call.
- Configuration over hardcoding. Routing rules, domain lists, stage definitions, lead days, brand voice, assignment logic, all of it is tenant scoped and database driven. If changing a client's behavior requires a code commit, that is a bug. This one is not negotiable, it is the reason the last architecture got reset.
- Integrate external systems properly. Gmail delegation model with Pub/Sub watch and renewal, OAuth token refresh and reconnection flows that survive real world token expiry, CRM handoff endpoints with strict validation at the boundary.
- Ship safely. Integration and unit tests on the paths that actually matter (merge logic, access control, client to case to document linkage, stage scheduling math). Reversible SQL migrations with backfills against live data. Deploy pipelines that fail loudly instead of silently.

What we are looking for

- 5+ years of full stack engineering , with meaningful time on TypeScript/Node, Postgres, and React (Next.js strongly preferred).
- You have written the code. Not managed it, not prompted it, not reviewed it from a distance. We will ask you to walk us through a system you built,



why the schema looks the way it does, and what you would change now. Depth of answer matters more than logo on the resume.
- Production database competence . Schema design, foreign keys and provenance modeling, migrations and backfills on live data with users on it, multi tenant scoping discipline, and an instinct for what a missing index or a leaky tenant filter costs you.
- AI native development workflow . You use Claude Code, Cursor, Copilot or similar every day to multiply output, and you retain full ownership of correctness. You review, test, and understand everything you ship. We will notice quickly if you cannot explain your own diff.
- Systems thinking in correctness sensitive domains. Role based access control, visibility and sharing models, dedup and merge, scheduling and date math, or similar work where being subtly wrong is worse than being slow.
- Strong written and spoken English, and comfort presenting to and debugging live with US clients.
- Bias toward root cause fixes. When you see a hardcoded value, a one off patch, or a silently swallowed error, you replace it with something configurable, observable, and tested.

Nice to have

- Gmail and Google Workspace API integrations , especially OAuth, delegation, and Pub/Sub watch and renewal.
- Prompt engineering for structured LLM output (classification, extraction, action generation) against the Anthropic Claude API .
- Supabase in production, including RLS, edge functions, and its rough edges.
- Docker and CI/CD pipelines, DigitalOcean deployment.
- Exposure to HR tech, payroll, service desks, or ticketing and case management.
- Comfort with a client delivery layer such as Rocketlane, and with UAT gated milestones.

📌 Product Engineer (Gurugram)
🏢 Clarwiz
📍 Gurugram