The biggest ideas from
all in one place.
read the recap
Atomicwork is building agentic service management for modern IT teams. Atom, our AI assistant, answers and resolves employee requests where people already work, and our service desk gives IT teams one place to run everything behind it. The hard design problems here are new ones: how an agent shows its work, when it should ask a human, and how an IT team trusts software that acts on its own.
We are looking for a designer who builds. You will own product surfaces end to end: frame the problem, explore in whichever medium gets you to the truth fastest, prototype in code, and ship to production with engineers. There is no handoff file and no walking away after the mock is approved.
This is an individual contributor role. You will not manage people. Your influence comes from the work you ship and the bar you set for everyone around you.
Most work starts in a conversation with an engineer and a PM. It moves quickly into a working prototype, often built with Claude Code or Cursor on top of our real React components. We put it in front of customers, then refine it for system consistency and polish. Figma is where we think about systems and detail, not the only place design happens. You will help shape this process as the tools keep changing.
Agent experiences that earn trust: showing what Atom is doing and why, how sure it is, when it hands over to a person, and how to undo what it did.
Dense, operational screens for IT teams (queues, request details, approvals, workflows) that stay calm and legible at scale.
Our design system, in Figma and in code, with the frontend team.
Prototypes wired to real models and real data, good enough to test with customers and to critique honestly.
Pull requests to our frontend for polish, interactions and UI fixes, reviewed like any other engineering change.
Quality after launch. You will use what you shipped, notice the paper cuts, and fix them.
Around 4–7 years designing software products, much of it in B2B SaaS. Years are a guide. What you have shipped matters more.
Excellent visual and interaction craft. Type, hierarchy, spacing and motion, and the small details too: focus states, empty states, the words on a button.
You build. You write solid HTML and CSS and can find your way around an existing frontend codebase. We build in React, so knowing it helps. If you don't, we expect you to pick it up from what you already know. You don't need to be a frontend engineer, but you should be able to ship UI polish yourself.
AI is part of how you actually work. You use tools like Claude Code, Cursor or Claude Design to explore, prototype and build, and you can explain, test and take responsibility for what they produce. You also know where they fall short and when not to use them.
Judgment over output. You can tell a generated screen from a solved problem, say what a prototype did not prove, and argue for killing a feature when it is the right call.
A working understanding of how LLM products behave (latency, uncertainty, mistakes, non-deterministic output) and what that means for the interface.
You are comfortable with ambiguity and with opinionated teammates. You hold strong opinions loosely and change your mind when the evidence says so.
You have read our values (agency, taste, ownership, mastery, impatience and customer obsession) and they sound like how you already work.
You have built or contributed to a design system in code: tokens, components, documentation.
You have designed conversational or agentic products, such as assistants in Slack or Teams, or copilots.
Something you made on your own: a side project, an open-source component, a tool your team still uses.
Experience with IT, HR or operations software.
React experience.
