Agentic Systems Architect

Description

The company is a global digital solutions and transformation business, and this role sits in its Cloud & Data Engineering function.

You will build end-to-end agentic systems that connect data, analytics and operations.

The work pairs agent infrastructure (LLMs, tools, memory, MCP) with data architecture (lakehouses, semantic layers, knowledge graphs) so reasoning over complex data stays safe, governed and cost-controlled.

You will work alongside talented colleagues, with room for autonomy, ownership and knowledge sharing.

Candidates who do not match every requirement are still encouraged to apply.

Job code: LI-MS2

Responsibilities

• Design multi-agent orchestration, such as orchestrator-worker and hierarchical patterns, and justify the architecture for each business use case.
• Build constrained agents with explicit guardrails, scoring gates and approval workflows for high-impact actions.
• Architect MCP servers for enterprise integration, defining authentication, role-based access, and stateful or stateless patterns.
• Create metric, dimension and business logic layers on Databricks.
• Implement text-to-SQL with cost estimation and query safeguards.
• Design knowledge graphs or ontologies for supplier and parts relationships, planning entity resolution and multi-hop queries.
• Choose a vector database (OpenSearch, Pinecone or native Databricks) and design chunking, hybrid retrieval and metadata filtering.
• Set up tracing for agents covering inputs, reasoning, tool calls, latency, tokens, cost and model versions.
• Build evaluation frameworks that act as release gates, using golden sets, LLM-as-judge and regression tests.
• Track cost per decision, alert on runaway loops and enforce per-tenant budgets.
• Design secure deployment across identity, secrets and compute, plus versioning, canary rollouts and on-call runbooks.

Requirements

• Required: 10+ years of experience across software engineering, platform engineering or systems architecture.
• Required: 2+ years shipping production LLM or agent applications, not prototypes.
• Required: hands-on work with multi-agent frameworks, including loops, tool calling, memory and constraints.
• Required: experience building and deploying MCP servers, including transport, authentication and statefulness.
• Required: experience shipping evaluation frameworks such as golden sets and LLM-as-judge as release gates, and measuring agent quality in production.
• Required: experience designing or working with a lakehouse on AWS, Databricks or similar, including semantic layers, text-to-SQL and governance.
• Required: production experience with vector databases, with knowledge of ANN, HNSW and chunking trade-offs, or experience building or maintaining knowledge graphs (RDF/OWL, Neo4j, property graphs) with entity resolution and relationship inference.
• Required: experience building observability for multi-step workflows, tracing tokens, latency, cost and model versions with tools such as OpenTelemetry or Dynatrace.
• Required: FinOps or cost optimization in cloud systems, with a record of controlling runaway compute or API costs.
• Required: Python, REST API design, infrastructure as code with Terraform, CI/CD, and AWS cloud infrastructure.
• Hands-on Databricks work with Genie, Mosaic AI Agent Framework and Unity Catalog.
• Automotive or supply-chain domain expertise.
• GraphRAG or multi-hop reasoning patterns.
• Model routing and cost optimization, such as choosing between Haiku and Opus.
• Human-in-the-loop approval workflows.
• Prompt versioning and systematic evaluation.

Conditions

Flexible working arrangements are offered, depending on role and location.

The company welcomes applicants regardless of protected characteristics and will provide reasonable accommodations during recruitment on request.

Benefits

• Industry-leading benefits, support and services focused on holistic health and wellbeing.