Sr. SW AI Engineer

About Us Visa is a world leader in payments technology, facilitating transactions between consumers, merchants, financial institutions and government entities across more than 200 countries and territories, dedicated to uplifting everyone, everywhere by being the best way to pay and be paid. At Visa, you'll have the opportunity to create impact at scale — tackling meaningful challenges, growing your skills and seeing your contributions impact lives around the world. Join Visa and do work that matters – to you, to your community, and to the world. Progress starts with you. Job Description Software Development Engineers are expert problem-solvers and builders who design, implement, and improve software applications and systems. In this role, engineers spend approximately 60–75% of their time coding, working hands-on with code, data, and modern tools (including AI-assisted development, cloud services, and automation frameworks) to deliver secure, scalable, and high-quality technology solutions that drive business outcomes in the fintech sector. They collaborate with cross-functional teams – product managers, designers, data scientists, QA, operations, and compliance – to translate business requirements into robust technical solutions, all while adhering to best practices, security standards, and regulatory requirements. All roles require digital fluency, including the ability to work with emerging technologies such as Generative AI tools (e.g. ChatGPT, Microsoft Copilot) to support everyday work. This role sits within an AI-first engineering team building production-grade agentic systems powered by large language models (LLMs). Engineers design and implement multi-agent orchestration frameworks, retrieval-augmented generation (RAG) pipelines, tool-calling integrations, and LLM inference services — all at enterprise scale on on-premise infrastructure. Key Responsibilities: Collaborate with stakeholders to interpret requirements for project components and incorporate feedback into future designs or solution fixes. Translate business requirements by verifying assumptions and escalating potential design issues to appropriate stakeholders. Participate in system design and architecture, refining code plans and contributing to design documentation. Contribute to project estimation and escalate issues that may cause delays. Develop, implement, and maintain code for products, services, or components, applying coding patterns, guidelines, and best practices. Apply debugging tools to verify assumptions and proactively flag issues before they occur. Participate in code reviews to ensure coding standards are followed and address routine pull requests. Create test plans and configure testing procedures to identify and resolve defects for multiple features. Respond to support requests during on-call rotations, troubleshooting issues and deploying fixes under guidance. Leverage and build knowledge of software developer tools to create, debug, and maintain code for components. Remain current in skills by investing time in training resources to improve product availability, reliability, efficiency, observability, and performance. AI & Agentic Systems Design, implement, and tune   multi-agent LLM workflows   using agent orchestration frameworks (e.g., LangGraph, LangGraph4j, LangChain4j), including declarative graph compilation, conditional routing between agent nodes, state persistence, and dynamic agent spawning at runtime. Build and maintain   ReAct (Reasoning + Acting) agent loops   — iterative Reason→Act→Observe cycles in which an LLM selects tools, interprets tool results, and continues until a terminal answer is produced. Implement   Supervisor/Orchestrator agent patterns   where a Lead Agent dynamically plans, spawns, and coordinates Specialist Sub-Agents using conditional graph edges and tool frameworks. Develop   MCP (Model Context Protocol) tool servers   that expose enterprise data sources as callable tools for AI agents, enabling natural-language-driven access to structured backend systems. Implement and refine   RAG (Retrieval-Augmented Generation) pipelines: document ingestion and chunking (PDF/Word), embedding generation, vector storage (e.g., PostgreSQL + pgvector), and semantic similarity retrieval to ground LLM responses in authoritative source documents. Integrate with   enterprise LLM inference gateways   via REST APIs, including request shaping, prompt template management, token budget control, and graceful degradation under service unavailability. Design and implement   agent memory architectures   — both short-term per-agent isolated context stores and long-term shared knowledge repositories used across agent workflows. Build   confidence scoring and uncertainty quantification   mechanisms for AI-generated outputs, including threshold-based routing of low-confidence results to human review. Implement   agent feedback loops   — capturing human reviewer decisions and feeding structured corrections back into prompt templates and knowledge repositories to continuously improve model behavior. Write   LangChain4j / LangChain tool definitions   (e.g., Java   @Tool-annotated methods) and wire them into ReAct agents for data transformation, external API calls, and domain-specific processing. Build and maintain   agentic audit trails   — structured logs capturing every tool call, agent decision, LLM prompt/response, and state transition for traceability and explainability. Develop and test   Spring Boot microservices   (Java 21, Spring Boot 3.5.x) that host AI orchestration engines, expose REST APIs for downstream consumers, and integrate with upstream data source systems. Author and maintain   Helm charts   and   Jenkins CI/CD pipelines   for containerized deployment of AI services to Kubernetes/OpenShift clusters across multi-region on-premise data centers. Build   end-to-end and integration tests   for non-deterministic AI components — including prompt