Featured work
Three engagements, three different problems
Continuous discovery
They needed the product before they knew what it should be.
A voice-AI product
A startup needed to work out what a realtime voice product should be before committing to it. In three weeks they had a speaking prototype with live evaluations; in six, an MVP on scalable cloud infrastructure. Two pilots followed, and they kept us on.
- 3 wks
- to a speaking prototype
Learning the domain
A paper process, no technology team, thousands of children.
A clinical assessment, rebuilt as software
An organisation with no in-house technology function. We learned the clinical domain and turned a paper developmental assessment for young children into a digital tool, handling FERPA-grade data, with the foundation laid so AI features can follow. It is in production today, and its customers pay per child on it.
- 2,600+
- children on the platform
Expertise and leadership
Six weeks, six engineers, no working agent.
A working agent in three days, leading their engineers through the build
A security startup’s seven engineers spent six weeks trying and failing to build a conversational agent. We built it with them in three days: the foundation on day one, ourselves; then, with the CTO, we split the remaining work, matching each piece to one of their engineers. By the afternoon of the third day, they demo’d the working app to a major crypto exchange and came away with a preliminary commitment.
- 3 days
- to a working agent