Senior AI Platform Architect: Multi-Agent Systems

Chubb; Toronto; On-site; CAD 150,000 - 190,000; Full time

Job summary

Chubb AI builds an enterprise AI platform and seeks an engineer to design and implement the core agent infrastructure, including orchestration engines, desktop harness, tool registries, and policy layers. You will work with frontier SDKs and agent loops, owning model-agnostic routing and grounding components while ensuring safe, scalable operations.

The role emphasizes hands-on development, architecture ownership, and collaboration with platform engineering, DevOps, security, and Model Risk

Qualifications

• 7+ years building production software with recent work on LLM-based systems.
• Strong TypeScript and Python (FastAPI, async, Pydantic).
• Production depth in Microsoft Agent Framework, LangGraph or Deep Agents.
• Direct experience with frontier provider SDK at the agent-loop level.
• Hands-on with tool calling, MCP or similar protocol, streaming (SSE/WebSocket), and structured output.
• Cloud-native delivery: containers, Kubernetes (AKS preferred), CI/CD, and observability.
• Ability to work through ambiguity with stakeholders.

Responsibilities

• Design and build core agent infrastructure: orchestration engines, desktop harness, and policy layers.
• Develop integration with frontier SDKs (Claude, OpenAI, Gemini) and agent SDKs.
• Own model-agnostic routing, cost accounting, and degradation strategies.
• Create tool registries, skill systems, and versioned artifacts for capabilities.
• Build retrieval, grounding, vector search, and citation pipelines.
• Develop offline eval, regression gates, and diagnosable failure tracing.
• Implement multi-runtime harness behind a single API and event stream.
• Create adapter layers for SDKs, events, and permission semantics; support runner/server split.
• Integrate MCP: stdio and remote SSE servers, and UI bridge for sandboxed widgets.
• Implement permission system (ALLOW/ASK/DENY) driven by policy hooks.
• Work on sandboxing: tokens, namespaces, Seatbelt/Bubblewrap, and container options.
• Schedule memory and continuity features: recurring tasks, cross-conversation references, indexing.
• Apply patterns: facade, adapters, factories, DI, registries.
• Maintain clear separation between business logic and infrastructure, with strong error handling.
• Ensure observability with logging, tracing, and memory-growth monitoring.
• Design multi-tenant architectures with entitlements and OAuth delegation.
• Write design docs.

Skills

LLM systems

TypeScript

Python

Microsoft Agent Framework

LangGraph

Deep Agents

Tool calling

MCP protocol

Streaming (SSE/WebSocket)

CI/CD

Observability

Cloud-native

Tools

FastAPI

Pydantic

Kubernetes

AKS

WebSocket

Job description

Chubb AI builds an enterprise AI platform and seeks an engineer to design and implement the core agent infrastructure, including orchestration engines, desktop harness, tool registries, and policy layers. You will work with frontier SDKs and agent loops, owning model-agnostic routing and grounding components while ensuring safe, scalable operations.

The role emphasizes hands-on development, architecture ownership, and collaboration with platform engineering, DevOps, security, and Model Risk