Headquarters: Toronto, Ontario, Canada
URL: https://toyo.ai
Senior AI Engineer — Toyo AI
Location: Remote (Eastern time zone preferred) · Full-time
Stage: Seed (CAD $6M raised) · 7 people
What we’re building
Small businesses are the engine of every economy. The founders running them started with a vision for where they wanted to take it. Most days, they never get there. Their time goes to working in the business rather than on it: context-switching all day between meetings, email, and an average stack of 30 different SaaS tools, with less than 30 minutes left for anything strategic. Software was supposed to help. Managing the chaos became its own full-time job. We built Toyo to fix that for ourselves first, and now we want to give it to everyone.
Toyo gives founders and their teams a platform where agents help run the entire business: personal workflows, sales, marketing, operations, and finance. Agents that know your business, work continuously, and reach out via WhatsApp, SMS, iMessage, or voice when something needs a human call.
We’re building the underlying primitives to make this real: isolated compute environments, persistent databases, durable objects, and a credential layer that lets agents take consequential actions without creating risk. OpenClaw showed what’s possible when an agent gets a full computer and persistent memory. We’re building that experience for a business owner who needs to trust it with their company, not a developer running it on their own machine. Security is in the architecture from day one.
We run it ourselves in production. A team of seven, operating at the output of a company three times our size. We’ve raised CAD $6M from Frontline Ventures and iNovia Capital. The founding team has previously built and sold companies to Workday, MessageBird, and Fox. We’re looking for a senior AI engineer to help build the infrastructure that makes this work at scale.
The role
You’ll be our Founding AI Engineer, joining a team of seven and shaping how we build and deploy agent infrastructure as we scale from early customers to a real platform. This is a senior individual contributor role. You’ll write a lot of code, make architectural decisions that matter, and work directly with the founding team on the things that are hardest to get right.
The work spans two areas: the agent systems themselves (how sessions run, how agents communicate, how the VM environment executes work) and the AI practices that govern how we build everything else, including agents-in-the-loop for our own development, evals, and reliability at scale.
The stack
The architecture has two layers working in tandem: an edge-native control plane handling real-time state, data, and API, and isolated secure VMs where agents actually execute, each with a full filesystem, persistent databases, and browser access.
Edge-native control plane: API routing and real-time session management run at the edge. We use Durable Objects extensively for stateful real-time primitives: per-org session registries, per-session message history and streaming, live voice AI call handling, and multi-channel messaging bridges for SMS, WhatsApp, and Slack. Long-running durable workflows handle operations that need to survive failures. A relational database sits behind a connection proxy for core application data.
Secure agent VMs: Agents run in isolated VMs, one per organisation. This is where Claude Code executes, with a full computer, filesystem, and browser. Sessions connect over ACP (Agent Client Protocol, JSON-RPC 2.0) via WebSocket. VM lifecycle, boot environments, snapshotting, and OS-level failure modes are central to this role.
Application layer: TypeScript throughout. The API layer is built on Hono, with full end-to-end type inference from server route definitions to the frontend client, no codegen, no separate schema. React 19 on the frontend. Authentication runs through an enterprise SSO layer and model routing goes through a provider-agnostic API, so we can swap and mix models without touching agent logic.
What you’ll work on
Agent reliability. Agents make mistakes. You’ll build the systems that make them trustworthy enough to do real work autonomously: human-in-the-loop checkpoints, tiered approval flows, and structured output validation that catches failures before they reach users.
Session infrastructure. ACP session lifecycle, resume after idle timeouts, multi-agent orchestration within a VM. The work is invisible when it’s done right, and users notice immediately when it isn’t.
Secure VM environments. Each organisation runs its own isolated VM with a full computer, filesystem, and browser. You’ll make that environment robust through boot orchestration, snapshotting, security hardening, and OS-level failure recovery.
AI-native development practices. We use agents to build Toyo. You’ll help define what that looks like in practice: automated feedback pipelines, agents that triage and enrich issues before a human touches them, and evals that tell us when something has actually improved.
Scheduled and event-driven agents. Cron-triggered agents, webhook-triggered sessions, WhatsApp and Slack integrations. The system should be doing useful work at 3am without anyone asking it to.
What we’re looking for
You’ve restructured how you work around AI, not just adopted it. Claude Code is where you spend most of your day. You follow the field closely enough to have real opinions, you know what works in production versus what demos well, and you try new techniques as soon as they’re available.
We don’t care where you went to school, or if you did. We care what you’ve built.
Strong signal:
- Claude Code is your primary development environment, not a tool you occasionally open
- You’ve shipped production agent systems and have scars from it
- You work with LLM APIs directly: prompt design, tool use, structured outputs, evals
- You understand edge-native and stateful distributed systems well enough to design around their constraints
- You’ve worked at the VM or container level and know what happens when things go wrong at the OS layer
- You write TypeScript well and care about correctness
Also useful:
- Experience with agent communication protocols (ACP, MCP, or similar)
- Background with WebSocket-heavy real-time systems
- You’ve thought seriously about agent security: credential isolation, output sanitisation, prompt injection
Not a fit if:
- You need a mature codebase and clear specs before you can move
- You think of AI as a productivity tool rather than a fundamental shift in how software gets built
What you get
- $200K-$325K CAD salary, benefits, and equity commensurate with a founding engineer role
- Direct influence on product and architecture from day one
- A small team that operates the way we’re selling: agents do the work, humans make the calls
- Remote-first, no bureaucratic overhead, and people who fix problems instead of scheduling meetings about them
- Once a year we take the whole team somewhere and spend a week on how we work together. Last year it was Sevilla. We’re already planning the next one.
How to apply
Email hello@toyo.ai with the subject line: Senior AI Engineer
Skip the cover letter. Send us something you’ve built: a repo, a write-up, a deployed project. Tell us what you think is hard about building reliable agent systems and how you’d approach it. We’ll read everything.
Requirements
- 10+ years software experience with production systems
- 5 years+ experience with Typescript, Node.js, React (front-end and back-end)
- Deployed production AI products/platforms
Benefits
- $200K-$325K CAD salary, benefits, and equity commensurate with a founding engineer role
- Comprehensive healthcare coverage for you and your family
- 5 weeks vacation – We actually expect you to take it
- Remote by default – Our team is scattered across time zones. No commutes. No headquarters. We meet in person when it matters.
- Annual team retreat – Once a year we take the whole team somewhere and spend a week working on how we work together. Last year it was Sevilla
To apply: https://weworkremotely.com/remote-jobs/toyo-ai-senior-full-stack-ai-engineer
Share: