AI Engineer

Payoneer; Gurugram District; On-site; INR 1,800,000 - 3,600,000; Full time

Job summary

Payoneer is seeking a hands-on AI systems engineer to own end-to-end production AI deployments. You will define problems, architect solutions, and ship with rigorous monitoring and observability in a fast-paced fintech setting.

You will collaborate with an AI Pod, design decision logic and success metrics, and ensure measurable outcomes. Expect fast iteration, strong evaluation, and a bias toward robust, scalable architectures.

Qualifications

• Experience shipping production AI systems in real-world environments.
• Fluency with AI coding tools and rapid prototyping.
• Strong evaluation instincts: define metrics, build test sets, and measure before shipping.
• Understanding of agentic architectures: orchestration, multi-step workflows, fallback and error logic.
• Comfortable across stacks to select the right solution.
• Honest communication: flag problems early and provide direct feedback.
• Fintech or regulated environments exposure is a bonus; unknown risks are not acceptable when money is moving.
• Curious: explore new models/tools and understand how things actually work.
• Innovative: run experiments, share learnings, push boundaries.
• Think in workflows and agentic loops; not just CRUD with an LLM bolted on.
• Production mindset: consider 2am incidents, cost risks, and edge cases from day one.
• Full-stack ownership: prompt, RAG pipeline, orchestrator, and backend involvement.
• Builder's mindset: prototype improvements, not accepting 'too hard to fix'.
• Grounded confidence: act with incomplete information when needed and own decisions.

Responsibilities

• Own the full loop: understand the business problem, define the solution, architect it, ship it, and monitor it.
• Build with AI as your primary tool using Claude Code, Cursor, and similar tools for rapid leverage.
• Co-design with your AI Pod to define decision logic, risk thresholds, and success metrics.
• Build the harness before shipping; ensure evaluation frameworks are part of the architecture.
• Own production: monitoring, alerting, observability, and on-call readiness.
• Raise the bar: contribute improvements in design reviews, code, and team processes.

Job description

Responsibilities:

• Own the full loop. Understand the business problem, define the solution, architect it, ship it, monitor it. No one hands you a spec.
• Build with AI as your primary tool. Claude Code, Cursor, whatever gives you the most leverage. You should be able to do in a day what used to take a team a week.
• Co-design with your AI Pod. Work directly with your pod and the business to define decision logic, risk thresholds, and success metrics. You're building capabilities, not features.
• Build the harness before you ship. You don't call something done if you can't measure it. Evaluation frameworks aren't optional; they're part of the architecture.
• Own production. Monitoring, alerting, observability these go in on day one, not after the first incident. You're on call for what you build.
• Raise the bar. When you see a better way, say so. In design reviews, in code, in how the team works.

Requirements:

• Experience shipping production AI systems, not just building them, but watching them run in the real world and fixing what breaks.
• Real fluency with AI coding tools, not "familiar with, " genuinely fast with.
• Strong eval instincts: you define metrics, build test sets, and don't ship until you can measure.
• Understanding of agentic architectures: orchestration, multi-step workflows, fallback and error logic.
• Comfortable across any stack you go where the right solution is.
• Honest communication: you flag problems early, give direct feedback, and say what you actually think.
• Fintech or regulated-environment experience is a bonus; what matters is understanding why "probably fine" isn't acceptable when money is moving.
• Curious. You explore new models and tools before anyone tells you to. You want to understand how things actually work, not just that they work.
• Innovative. You don't wait for best practices to exist; you run the experiment and help create them. You iterate fast, share what you learn, and push the boundaries of what's possible.
• Thinks in workflows, not endpoints. You naturally decompose problems into agentic loops, skills, and orchestrated flows. CRUD with an LLM bolted on isn't a system you know; the difference.
• You feel the weight of production. When you're designing, you're already thinking about 2 am incidents, runaway inference costs, and edge cases that only surface when real money is moving. That instinct to stress-test your own assumptions before shipping is just how you think.
• Full-stack owner. You're comfortable touching the prompt, the RAG pipeline, the orchestrator, and the backend. You move up and down the stack to get the job done.
• Builder's itch. When you see a repetitive task, or a problem the team has just accepted as "too hard to fix, ", you can't leave it alone. You go home thinking about it. You come back with a working prototype. Watching people do manually what a system could do makes you uncomfortable in a productive way.
• Grounded confidence. You trust your abilities. You don't need sign-off before every decision. You move forward with incomplete information, make a call, and own it. And when you're wrong, you update and move on no drama, no ego. The gap between where you are and where you want to be is what gets you out of bed.