Forward Deployed Engineer at Razorpay

Razorpay, the Bengaluru payments company, hires Forward Deployed Engineers to ship AI agents inside its merchants’ businesses. It is one of the few Indian product companies using the FDE title for a role that looks like the original model: you sit with the customer’s engineers, own the integration end to end and stay on until it works in production.

Why Razorpay needs FDEs

At its FTX’26 event in March 2026, Razorpay launched Agent Studio, a marketplace for AI agents in payments and banking (Razorpay’s own blog lists the no-code agent builder as a beta), according to The Paypers. Launch agents included abandoned-cart conversion over voice or messaging, dispute response, subscription recovery (built with ElevenLabs) and cash-flow forecasting, with integrations into Shopify, WhatsApp, Shiprocket, Slack and QuickBooks. The press release says Agent Studio is built with Anthropic’s Claude Agent SDK.

Off-the-shelf agents only go so far for large merchants. The FDE is the engineer who makes them fit a specific merchant’s systems.

What the FDE owns

Razorpay’s Bengaluru listing describes the full loop:

• Pre-sale discovery. Assess integration complexity and whether a workflow suits an agent, then co-author the deployment plan in the proposal.
• Integration lifecycle. Work with merchant engineering teams from architecture to go-live.
• Agent building. Tool bindings, MCP integrations, orchestration pipelines, evals and human-approval flows on Agent Studio.
• Cross-system debugging. Across Razorpay’s systems, the merchant’s systems, external tools and model behaviour.
• Hypercare. First responder for 15–30 days after each go-live.
• Making it repeatable. Quarterly deployment blueprints for the shared knowledge base, and structured feedback to product managers.

The listing is explicit that this is embedded engineering, not pre-sales, support or solutions architecture. If you are a solutions engineer, that line matters: you will need to show you write and debug production code, not only design it.

Experience, skills and pay

Pay for the FDE role is not published. For context, FACE Prep cites aggregator estimates of ₹25–45 LPA for Razorpay’s ML Engineer role and ₹14–20 LPA for SDE-1, and calls them directional. Those are different roles, so use them only as a sense of Razorpay’s pay levels. For FDE pay at Indian product companies with 3–6 years, compare with the FDE Salary Estimator.

How to prepare

1. Ship a payments-flavoured agent. For example, an agent that reads a webhook for a failed payment, checks order data through an API and sends a WhatsApp recovery message, with a human approval step before any refund.
2. Show debugging across boundaries. Have one story where the bug was in someone else’s system and you found it through logs, retries or a webhook signature mismatch.
3. Learn MCP and one agent framework properly. The listing names MCP, LangChain, LangGraph and AutoGen. Build tool bindings and an eval set, not just a chat demo.
4. Practise discovery. Given a merchant problem, can you say in 10 minutes whether it suits an agent, what it integrates with and what could go wrong? That is the pre-sale part of the job.
5. Bring a hypercare mindset. Talk about monitoring, rollback and what you did in the first two weeks after something you built went live.

If you are moving from backend engineering, our guide on how the FDE role works in product companies explains what changes day to day.

Razorpay FDE interview: what is publicly known

1. Process for the FDE role: Razorpay's FDE listing does not describe the interview rounds.
2. What to expect from the job description: The role covers pre-sale technical discovery, merchant integration, agent building with evals and production debugging, so expect to be tested on reading unfamiliar code, API and webhook integration, an agent or LLM design discussion and how you work with external technical teams.