Senior Software Engineer, Applied AI | On-site | George WC

About the Role

• ARC AI builds AI agents as products: systems that operate autonomously across business functions, handle real customer interactions, and deliver measurable outcomes in a regulated insurance environment. This is not an experimentation function. We ship.
• We are looking for a Senior Software Engineer who has built and operated AI agent systems in production. Someone who knows what it takes to move an agent from prototype to live deployment, and who understands that evaluation, not iteration speed, separates reliable AI products from expensive demos.

Requirements

What You Will Own

• You will design, build, and operate the AI agent systems that power ARC AI products. This means owning the full arc: architecture, evaluation pipelines, production reliability, guardrails, and the backend integrations that connect agents to real enterprise data.
• You will set the engineering standards the rest of the team follows. Evaluation is a first-class responsibility here, not a QA afterthought. Securing an agent means architecting enforcement outside the model; defences sit at the execution boundary, not inside the system prompt. Compliance with POPIA and GDPR is an architectural decision you make at the start, not a checkpoint at the end.
• The architectural decisions you make in this role will be the ones every subsequent engineer builds on.

What We Require

Non-negotiable

• Demonstrated experience building and shipping AI agent systems in production: not demos, not internal tools that never went live
• Ownership of an evaluation pipeline for a production AI system: you defined the metrics, built the framework, and used it to make deployment decisions
• Experience debugging production AI failures: you have traced a silent agent degradation to its root cause in a live system
• Proficiency in Python and LLM frameworks LangChain, LlamaIndex, or equivalent
• RAG architecture: retrieval pipeline design, vector database implementation, chunking and embedding strategy
• API design and backend integration at production scale
• Secure tool access patterns for agentic systems: you know where LLM reasoning ends and deterministic enforcement must begin, and you have built the boundary between them
• Experience implementing guardrails: input validation, output filtering, and execution-layer prompt injection defence
• CI/CD for agentic systems: you have implemented progressive delivery pipelines for agents, including instrumentation of tool invocations and decision points, staging with regression benchmarks, and canary deployment to detect behavioural drift before full rollout

Strong Advantage

• Experience in regulated financial services, insurance, or healthcare environments
• Familiarity with graph-based agentic orchestration frameworks LangGraph or equivalent, particularly durable execution and human-in-the-loop checkpointing in regulated environments
• Familiarity with Azure AI tooling and services
• MLOps practices: monitoring, observability, cost management for inference workloads
• CRM and enterprise system integration

Educational Requirements

• Bachelor's degree in Computer Science, Software Engineering, Information Technology, Artificial Intelligence, Data Science, or a related field
• OR equivalent practical industry experience building and deploying production-grade AI systems