Forward Deployed Engineer

Agent Studio helps Razorpay merchants run AI agents across payments, checkout, disputes, subscriptions, reconciliation, risk, and customer workflows. We take the messy, undocumented work merchants do every day and turn it into reliable production agents.

We're hiring a Forward-Deployed Engineer who is equal parts engineer, solution consultant, and merchant partner — the single technical owner of the merchant relationship. You'll sit directly with strategic merchants, run the discovery, find the real constraint behind the ask, design the solution, build the first working version, prove whether it moved the metric, and expand from there.

This is a deeply customer-facing role. You'll own the conversation — not hand it off. You'll walk into a room where the business owner, the ops lead, and the dashboard all tell different stories, and you'll be the one who reconciles them into a spec everyone agrees to ship. Then you'll go build it. You earn the customer's trust because you can do both — reshape the ask and write the code.

If you like being close to the problem, comfortable in front of customers, and can ship, this is for you.

• Own merchant conversations end-to-end — discovery calls, workshops, solution reviews, and rollout check-ins. You're the merchant's trusted technical advisor.
• Sit inside merchant workflows long enough to understand the real process, not the documented one. Ask who actually does the work and what happens when it fails.
• Reconcile conflicting signals across business owners, operators, dashboards, CRM, and ticket queues into a clear problem statement.

• Name the business constraint behind the ask: checkout recovery, dispute win rate, RTO loss, SIP recovery, settlement delay, refund spikes, support load, or cashflow leakage.
• Convert fuzzy merchant problems into crisp agent specs: success metric, data sources, tool calls, guardrails, human-in-the-loop points, and rollout plan.
• Decide what the agent does on its own, what it recommends, and where a human must approve.
• Write the spec clearly enough that product, engineering, legal, and the merchant all agree on what ships.

• Build prototypes, connectors, MCP tool specs, eval fixtures, dashboards, and the glue code that gets an agent into production.
• Debug integration failures across Razorpay systems, merchant systems, external tools, and model behavior.
• Deploy across the constraints that real merchants have — auth, networking, data access, latency, cost per run, observability, and compliance — and keep them in the design from day one.

• Set quality gates and outcome metrics before rollout. Track production outcomes, not demo applause.
• Measure whether the agent moved the target metric. If it didn't, diagnose and fix — or kill it with evidence.
• Drive expansion through building: spot the next high-value use case, prototype it, and deepen the merchant's agent footprint over time.
• Turn repeated field patterns into reusable Agent Studio capabilities — connectors, templates, eval packs, docs, platform requirements. You are the field voice that shapes what the platform builds next.

• Have 3+ years shipping production software, and are strong in Python or TypeScript.
• Have worked directly with customers, enterprise stakeholders, or operators — and are comfortable owning the room and the technical relationship.
• Have hands-on experience with APIs, webhooks, auth, data pipelines, and production debugging.
• Have practical LLM/agent experience: prompts, tool calling, RAG, evals, or workflow automation.
• Prefer ambiguous, high-context problems over clean tickets.
• Take ambiguous problems, define the structure, and drive them to resolution 
• Write clearly — specs, implementation plans, and crisp follow-ups.

• Prior shape: solutions engineer, solution consultant, sales engineer, founding engineer, product engineer, applied AI engineer, or an operator who wrote code.
• Payments, fintech, ecommerce, subscriptions, risk, reconciliation, settlement, or finance-ops experience.
• Worked with MCP servers, agent skills, multi-agent workflows, eval harnesses, or LLM observability.
• Enterprise integration experience: private connectors, VPC/on-prem constraints, SOC2/ISO reviews.

• Not a pre-sales demo role where success is a slick walkthrough.
• Not a pure implementation role where success is shipping whatever the merchant asked for.
• Not a support role where success is closing tickets.
• Not a research role where success is a clever agent that never reaches production.