Sr AI / Agentic Engineer

What matters most

• Agentic Workflow Orchestration

  The core of this role is designing and implementing AI agents capable of autonomous task execution. You must demonstrate proficiency in LangChain and LangGraph to build complex, multi-step workflows that move beyond simple chat interfaces into functional business process automation.

• Enterprise Data Grounding and Integration

  You are expected to bridge the gap between LLMs and enterprise data. This requires building RAG and semantic retrieval solutions that utilize enterprise ontologies to ensure AI responses are accurate, context-aware, and grounded in specific business knowledge rather than generic training data.

• Model Context Protocol (MCP) Implementation

  A specific technical requirement is the development and maintenance of MCP integrations. You must be able to connect disparate enterprise APIs and semantic services to AI agents, ensuring seamless data flow and interoperability within the organization's technical ecosystem.

Likely challenges to think through

• Balancing Autonomy with Reliability

  Developing autonomous agents for business workflows introduces the risk of unpredictable behavior. Prepare to discuss strategies for implementing guardrails, monitoring agent performance, and ensuring human-in-the-loop oversight for critical operational tasks.

• Managing Complex Semantic Architectures

  Integrating enterprise ontologies and knowledge graphs with vector databases is technically demanding. You may face challenges in maintaining data consistency and retrieval accuracy; be ready to explain how you handle schema mapping and semantic alignment at scale.

Topics to prepare

• Agentic Workflow Design

  Focus on the conceptual architecture of autonomous agents. Exercise: Map out a multi-step business process (e.g., automated procurement) and define the decision-making logic, tool-use requirements, and error-handling loops needed for an agent to execute it autonomously. This is directly relevant to the requirement for designing and implementing AI agents for business workflows.

• Enterprise Ontology and RAG Grounding

  Study how to map unstructured data to structured enterprise ontologies to reduce hallucinations. Exercise: Design a schema that links a specific domain's terminology to a vector database retrieval process. This addresses the requirement to ground LLM responses using enterprise ontology for accurate, actionable insights.

Skills to revise

• Model Context Protocol (MCP) Implementation

  Gain hands-on experience building MCP servers to expose local or remote data sources to LLMs. Exercise: Create a simple MCP server that connects a mock enterprise API to a client, allowing an LLM to query the API via the protocol. This is a core requirement for building and maintaining MCP integrations.

• LangGraph for Multi-Agent Orchestration

  Practice building stateful, multi-actor systems. Exercise: Build a LangGraph workflow where one agent acts as a researcher and another as a reviewer, passing state between them to refine an output. This is essential for the autonomous task execution and multi-agent system requirements.

• Vector Database Optimization

  Focus on retrieval performance and relevance. Exercise: Implement a hybrid search approach combining keyword-based retrieval with semantic vector search to improve the accuracy of a RAG pipeline. This is critical for developing the requested semantic retrieval solutions.

• Build a Multi-Agent Workflow Prototype

  Develop a small-scale project using LangGraph that demonstrates autonomous task execution. Focus on creating a system where multiple agents collaborate to complete a business workflow, such as automated document processing or data extraction, to showcase your ability to handle complex agentic logic.

• Implement an MCP Server Integration

  Create a functional Model Context Protocol (MCP) server that connects a local LLM or LangChain application to an external enterprise API or semantic service. Document the process of defining resources and tools within the MCP framework to demonstrate your technical proficiency with this specific protocol.

• Design a RAG System with Enterprise Ontology

  Build a Retrieval-Augmented Generation (RAG) pipeline that uses a vector database and incorporates a knowledge graph or structured ontology to ground LLM responses. Prepare a walkthrough explaining how you ensured factual accuracy and semantic relevance in the generated outputs.

• Prepare Technical Evidence for AWS Bedrock

  Review your experience with AWS Bedrock and prepare to discuss specific use cases where you deployed models or managed LLM integrations within an AWS environment. Be ready to explain how you optimized performance and security for enterprise-scale AI applications.

• Formulate Strategic Questions for the Interview

  Prepare insightful questions regarding NTT DATA's current approach to enterprise ontology, their specific challenges in scaling agentic workflows, and how they balance autonomous task execution with human-in-the-loop operational requirements.