Building products
Building an AI product across web, backend, and mobile
An AI feature lives inside a whole product. Web and mobile interfaces, backend services, and persistent state all shape how people use it.
The model is one component
It is tempting to describe an AI product by naming its model or framework. But a person uses a whole application: they need an interface to describe a goal, a backend to coordinate the work, and somewhere to keep track of what happened.
I build across those layers, from web and mobile interfaces to backend systems and agent orchestration. Each layer has a part in making an AI capability dependable and understandable.
Give every layer a clear job
For Crew, the web and mobile apps are ways to control and observe work. The control plane handles authentication, task state, orchestration, permissions, and observability. The execution side runs agents, tools, and computer-based work.
The mobile app brings the same control and visibility to a smaller screen. React Native and Expo make it possible to build that mobile interface alongside the web application while the backend and agent runtime handle the actual execution.
Separating those responsibilities helps the product grow without making either interface responsible for work that belongs in an agent runtime or remote environment.
Build for the complete task
A useful AI feature has to fit the user’s actual task. That may mean saving an artifact, delivering a digest, or returning a code review with clear findings. The surrounding software makes those outcomes possible.
Thinking in full-stack terms keeps me focused on the complete experience: not only what the model can say, but how the product helps someone get from a goal to a result.