SDE-2, FPG Engineering

About the Role

We are looking for a Software Engineer (SDE-II) with strong Applied AI fundamentals and an AI-Native engineering mindset to join Flipkart's Fintech & Payments Engineering Group (FPG).

FPG powers payments, credit, fraud detection, and financial infrastructure for hundreds of millions of Indians. You will build intelligent features and systems where AI is the primary lever, from LLM-powered workflows to ML-integrated decisioning. This is a role for engineers who believe AI fundamentally changes how software is built, not just what it does.

You will grow rapidly alongside some of the sharpest fintech and AI engineers in the country, solving problems at a scale.

Job Responsibilities

1. Agentic Development Lifecycle (ADLC)

• Build autonomous agent components that execute ADLC stages: agents that decompose engineering tasks, generate implementation plans, produce code, write test cases, and validate outputs within a larger multi-agent pipeline.
• Implement agent tool integrations for development lifecycle tasks: code execution environments, static analysis runners, test harness hooks, API schema validators, and CI/CD pipeline triggers that agents call autonomously.
• Build and operate agent evaluation pipelines: design test suites for agent tasks (does the agent produce correct, domain-appropriate code?), track agent performance over time, and identify where agents fail or produce unreliable outputs.
• Implement human-in-the-loop checkpoints: build the approval, review, and override interfaces that allow engineers to supervise agent-driven ADLC stages without becoming the bottleneck.
• Contribute to agent memory and context systems: retrieval pipelines that give agents access to codebase context, domain knowledge, API contracts, and historical engineering decisions.
• Operate production ADLC agents: monitor agent reliability, triage agent failures, tune prompts and tool configurations, and iterate rapidly on agent behaviour based on production feedback.

2. Domain Agent Engineering

• Build domain-specialised agents for fintech and payments tasks: agents that generate code aware of idempotency requirements, transaction integrity, and financial correctness.
• Implement validator agents and critic agents that review agent-generated fintech code against domain correctness rules before human review.
• Build test-generation agents for financial edge cases: payment failure scenarios, partial settlement, currency handling, duplicate detection, and fraud signal integration.

3. Agent Infrastructure Contribution

• Contribute to FPG's shared agent infrastructure: tool registries, agent orchestration components, shared memory stores, and evaluation pipelines used across teams.
• Build reliable backend services in Java and/or Python that underpin agent tool calls: low-latency APIs, idempotent endpoints, and observable services that agents can invoke safely at high frequency.
• Implement agent safety controls: rate limiting, action logging, and human-approval hooks for agent-initiated operations with production impact.

4. Growth & Engineering Rigour

• Develop genuine domain depth in fintech and payments - understand the correctness and compliance requirements that ADLC agents must satisfy to be trusted for production use in a regulated environment.
• Advance your agent engineering skills: multi-agent design patterns, agent evaluation methodology, reliability engineering for autonomous systems, and domain-specific agent specialisation.
• Take end-to-end ownership of the agent components you build - not just building them, but operating them, improving them, and measuring their real impact on development velocity and quality.

What Success Looks Like

• The ADLC agent components you build reliably automate development lifecycle tasks that previously required manual engineering effort with measurable quality and correctness.
• You evaluate agent outputs with the same rigour you would apply to human-written code: you know when an agent is right, when it is plausibly wrong, and when it needs a domain expert to review.
• You take full ownership of what you build from design through production operation without needing to be asked.
• You are visibly growing toward SDE-III: taking on more complex agent system design and beginning to define standards and patterns for the team.

You (Must Haves)

• 5+ years of software engineering experience, with demonstrated hands-on work building agent systems, not just using LLM APIs, but building systems where agents take autonomous actions.
• Experience building agents with real tool use: agents that execute code, call APIs, query databases, run tests, or perform other consequential actions.
• Familiarity with multi-agent patterns: task decomposition, agent handoffs, validator/critic agents, and managing agent state across multi-step workflows.
• Understanding of agent failure modes and how to detect and handle them: hallucination, tool misuse, context drift, and runaway execution.
• Strong distributed engineering fundamentals: APIs, databases, messaging, and service design.
• Proficiency in Java and Python, working experience with LLM APIs and agent frameworks
• BE/BTech in Computer Science or equivalent.

Nice to Haves

• Exposure to fintech, payments, or financial services - understanding the domain-specific correctness requirements that ADLC agents must satisfy.
• Experience building or contributing to code-generating agents, code-review agents, or test-generation agents.
• Familiarity with MCP (Model Context Protocol) - building tool surfaces or consuming them within agent pipelines.
• Experience with agent evaluation frameworks: designing benchmarks and test suites for autonomous agent task performance.
• Open-source contributions, technical writing, or community presence in agentic AI or developer tooling.