Building a product from scratch at an early-stage startup

The challenge

A behavioural intelligence startup was building a platform to help researchers and organisations make decisions based on real human behaviour, not self-reported data. When I joined, there was backend work underway but nothing on the design side. No brand, no website, no design system, no product UI. The target audience was still being defined, the team was small and fully remote across multiple time zones, and priorities were constantly shifting between investor needs, pilot programmes, and core product development.

The solution

I built the design foundation from the ground up. That started with the brand identity and logo, then a public-facing website, followed by the product UI for the app itself, where users create projects, conduct interviews, and generate reports. Alongside that, I set up a Figma design system, established a component library using shadcn, and connected it all to development through Storybook. Sprint planning happened in Linear to keep design and engineering aligned.

My role

I was the sole designer, working across brand, product, and frontend. I designed and built both the marketing site and the product interface, created the design system and component library, mapped and improved user journeys, designed dashboards for complex data, and worked directly alongside the developer using Claude Code and IDE tooling to assist with builds. Beyond design, I helped shape product direction, contributed to sprint planning, and adapted constantly as priorities shifted between investor demos, pilot needs, and long-term product goals.

The impact

The startup went from having no design presence to a fully branded product with a live website, a working app, and a scalable design system. The Figma library and Storybook setup means components are documented, consistent, and ready for any developer to pick up. The groundwork is in place for the product to scale as the team and user base grow.

Key takeaways

Working at this stage means there's no playbook. You figure it out, own it, and keep moving. The most valuable thing I've taken from this is learning how to build with constraints, pivot when needed, and treat every part of the product as connected.