Principal Engineer, Agentic Products

About the companythe company exists to make fans.

We are a profitable, independent Canadian gaming company focused on building, operating, and growing long-standing online game communities. Our flagship titles, YoWorld and FishWorld, have entertained millions of players over their lifetime. These are enduring live-service virtual worlds with rich in-game economies, virtual goods, deep social interaction, and player communities measured in years, not sessions.

We are now rebuilding the company as an AI-first studio. That is not a slogan here. It is an operating standard.

AI agents are already changing how we build, test, review, produce, and operate. We believe a small team of exceptional builders, amplified by well-architected AI systems, can outperform teams many times larger. This role exists to help prove that, operationalize it, and define what the next generation of game development looks like.

About the Rolethe company is hiring a Principal Engineer, Agentic Products & Workflows to build the systems that make an AI-first game studio real.

This is a founding-level engineering seat at the center of our AI transformation. You will help architect and build the agentic platform, autonomous engineering workflows, AI-enabled content pipelines, review systems, integration layers, and observability infrastructure that allow agents to produce reliable, reviewable, production-grade output for live games.

This is not a role for someone who is simply curious about AI.

You should already be building with agents, shipping systems, testing workflows, using AI coding tools daily, and developing opinions from real production experience. We care less about pedigree and more about what you have built, how you think, and whether your systems survive contact with real users, real content, real revenue, and real production constraints.

You will own verticals end to end: architecture, orchestration, data models, backend services, APIs, evaluation loops, human review surfaces, observability, quality controls, and the production workflows teams use every day.

This is hard in a very specific way. Agentic systems are nondeterministic, tool-driven, context-dependent, and often operating with minimal supervision. When they fail, they can fail quietly, expensively, or at the wrong moment. Your job is to build the infrastructure that makes those systems legible, reliable, recoverable, and trusted.

If that problem excites you, this role is built for you.

What You’ll Build

You will build and own systems across two flagship workstreams.

Autonomous Engineering Platform

The internal platform that lets a small number of engineers safely direct a large amount of autonomous work.

This includes:

• Orchestration primitives for multi-agent, multi-model workflows, including routing, scheduling, state management, memory, task decomposition, and recovery.
• Autonomous engineering workflows that can take a scoped ticket toward a reviewed, tested pull request.
• AI-to-AI review layers where agents critique, test, validate, and gate each other’s output before a human needs to intervene.
• Self-healing workflows with retries, escalation paths, rollback, recovery, and failure classification.
• Cross-agent memory and calibration systems so knowledge compounds instead of disappearing between runs.
• Prediction and confidence systems that help agents know when to proceed, when to ask, and when to stop.
• Evaluation infrastructure, test harnesses, traces, metrics, and dashboards that catch hallucinations, regressions, grounding gaps, quality drift, and cost spikes before they reach production.
• Integration layers that connect agents safely to codebases, tickets, repositories, data sources, tools, documentation, and production workflows.

LiveOps Content Pipelines

The AI-enabled production systems that turn creative direction into production-ready game content at scale.

This includes:

• Content generation workflows that produce structured, reviewable output aligned to YoWorld, FishWorld, and our quality standards.
• Human review surfaces where designers, artists, product managers, and live operations teams can steer, approve, reject, revise, and improve generated content.
• Prompt and spec composition systems that make quality repeatable rather than dependent on individual heroics.
• Audit, retry, versioning, and approval infrastructure so every input, output, revision, decision, and release step is tracked and recoverable.
• Tooling that helps creative and production teams move faster without lowering the bar.
• Integration layers that move approved content safely into live game workflows.

You will own the quality of what these systems produce, not just the plumbing