Swarm

The context

Swarm is the agent orchestration layer of PerkOS — a platform for deploying, managing, and scaling AI agents across organisations. It gives teams a real-time view of what every agent is doing, lets them assign tasks, monitor progress, and course-correct without touching code. I joined at the prototype stage, working closely with the engineering lead to take the product from a rough internal tool to a shippable desktop and mobile experience.

The problem

Swarm's early users came from web3 — comfortable with task boards and contributor management, but unfamiliar with AI orchestration. The core challenge wasn't building something new; it was making something genuinely complex feel like something they already knew. Every added layer of agent behaviour had the potential to confuse rather than empower, and the product needed to ship fast without sacrificing clarity.

What I did

• Reviewed the UX of the platform's first versions.
• Mapped analogous task management platforms (Trello, Jira, Notion) to ground the design in patterns AI and Web3 users already know.
• Studied how AI agents behave inside interfaces (Claude, OpenChat) to inform interaction patterns.
• Explored design directions and built a component system using Figma.
• Contributed to the architecture for managing agents and delivered a handoff-ready Figma prototype.

The result

The design reframes agents as team members and users as project managers. Once that mental model is established in onboarding, everything else — dashboards, task queues, agent profiles — feels intuitive. The interface scales from a single agent to a full fleet without changing its logic. PerkOS used the Figma prototype to ship the first internal version and to close their next funding round.