AI Application Engineer

What matters most

• Build grounded AI applications

  The posting calls for AI applications, agents and copilots using RAG, vector search, LLMs and AI orchestration across SAP and other enterprise platforms.

• Connect agents to enterprise systems

  The role integrates AI with SAP systems, ERP data and business applications; it also specifies hands-on agent development using MCP servers, A2A protocols and integration platforms.

• Engineer scalable services

  The role builds backend services, REST APIs and microservices. Strong Python proficiency is specified, alongside experience with JavaScript/TypeScript, Node.js or Java.

• Deliver production-ready use cases

  The posting emphasizes security, scalability, observability and performance, with cloud, containers and CI/CD exposure supporting production readiness.

• Align delivery with governance

  The role works with business stakeholders, product teams and architects on end-to-end use cases while implementing responsible AI practices, governance controls and compliance standards.

Likely challenges to think through

• Grounding versus response speed

  Preparation hypothesis: RAG and vector search may need to balance answer relevance against retrieval latency. Prepare to explain how you would evaluate retrieval quality, handle weak matches and measure end-to-end performance.

• Agent capability versus controlled access

  Preparation hypothesis: connecting agents to SAP and ERP systems through tools or protocols could make useful actions harder to constrain. Prepare an integration design covering authorization, limited tool permissions, validation and auditability.

• Rapid delivery versus operational rigor

  Preparation hypothesis: stakeholder-facing AI use cases may need iteration while still meeting production and governance expectations. Prepare to describe release criteria, monitoring, failure handling and responsible-AI checks without assuming a specific Infosys process.

Topics to prepare

• RAG and vector-search design

  Draw a RAG architecture for retrieving ERP records; explain indexing, retrieval, access boundaries and how you would check answer quality. Relevant because the role explicitly designs RAG architectures and vector search.

• Agent orchestration and tool use

  Map an agent workflow from prompt to function call to response, including failure and approval paths; explain where MCP servers and A2A protocols fit. Relevant to the stated agent-development and orchestration requirements.

• Enterprise integration architecture

  Sketch data flows between an AI service, SAP or another ERP source, and a business application; mark API boundaries and data permissions. Relevant because the role integrates AI solutions across enterprise platforms. Review SAP Integration Suite as a preferred platform, not a mandatory one.

• Production architecture trade-offs

  Compare two ways to scale an AI-backed microservice and describe what you would monitor for latency, errors and performance. Relevant to the posting’s scalability, observability and production-readiness responsibilities.

• Responsible AI and governance

  Review a hypothetical copilot response that exposes sensitive ERP data; identify security, governance and compliance controls that would prevent or detect it. Relevant to the role’s explicit responsible-AI and AI-security expectations.

Skills to revise

• Build a Python RAG prototype

  Using sample, non-sensitive enterprise records, implement retrieval and an LLM answer that cites its retrieved context; test a question with no matching record. This practices the required Python, LLM and RAG work. The sample prototype is a preparation exercise, not a stated employer project.

• Implement a controlled agent tool call

  Create a small agent that calls a read-only mock ERP function, validates its inputs and handles a failed call. This rehearses prompt engineering, function calling and AI-agent development specified in the posting.

• Develop an integration API

  Implement a REST endpoint that accepts a request, obtains data from a mock enterprise service and returns a structured response; test authorization and error handling. This targets the stated backend, API, microservice and enterprise-integration work.

• Practice a named AI framework and vector database

  Connect a small retrieval workflow in one listed framework, such as LangChain or LlamaIndex, to one listed vector database, such as Chroma or Qdrant; inspect a poor retrieval result. The posting asks for framework knowledge and vector-database experience, not every listed product.

• Package and check a service for release

  Containerize the practice API with Docker, run a basic CI/CD build and test, and explain how you would deploy and observe it in a cloud or Kubernetes environment. This addresses the posting’s cloud and DevOps exposure plus production-readiness expectations; the particular deployment setup is a preparation suggestion.

• Build an evidence map

  Prepare a factual one-page map of the posting’s 9–13 years and Bachelor of Engineering criteria, plus examples of Python, LLM, agent, RAG and enterprise-integration work. Mark SAP BTP and related SAP tools as preferred, not mandatory.

• Rehearse a project walkthrough

  Choose a real AI application or agent project, if available. Prepare a five-minute account of the business need, your decisions, stakeholder collaboration, results and trade-offs. Label any hypothetical example as a proposed design.

• Draw a RAG architecture

  As a practice exercise, sketch ingestion, vector search, retrieval, prompting and evaluation for an enterprise copilot. Explain how you would choose a vector database or orchestration framework rather than implying the posting mandates one.

• Design an agent-to-enterprise integration

  Prepare a diagram showing function calls, MCP servers or A2A interactions, a Python API or microservice, and access to SAP or other ERP data. Be ready to discuss permissions, failure handling and integration boundaries.

• Review production-readiness decisions

  Make a short checklist covering security, scalability, observability, performance and testing. Add how cloud deployment, Docker, Kubernetes and CI/CD could support those goals; these are discussion points, not a claimed Infosys process.

• Prepare a responsible-AI risk example

  Write down one plausible enterprise-agent failure, such as an unauthorized data retrieval or unsupported answer, and the governance, security and compliance controls you would propose.

• Draft questions for the employer

  Bring questions about the first AI use cases, SAP versus other platform integrations, how solution quality is evaluated, and who owns production and governance decisions.