Deca Labs
A cloud plan is cheap until the vendor cuts how much you can use, raises the price, or swaps the model. If that plan is running part of the business, the work stops with it.
We start from the jobs you need done, then pick the smallest computer that can run them. We set the model up on that machine. You keep the files, and you keep using it after the subscription changes.
We also build the software around that work: the apps, the search, and the sites. The engineers who usually do this are employed by large companies. We hire the ones who also take outside projects.
The studies below are software we have already shipped. Local AI is the same kind of job: a workflow, a computer that can run it, and software you can still use after we leave.
Study
Android and iOS apps for a US business bank, shipped in about six weeks.
A US bank for small and mid-size businesses needed phone apps, not another browser tab. Two engineers built Android and iOS together: balances, cards, and the chat assistant that was already running on the server.
Study
A website, a booking tool, and a small ad test for a chauffeur who had been living on referrals.
A private car service had no website and no booking system. We built the site, a chat that takes a booking, a phone app for the driver, and a short Facebook and Instagram test. In two weeks, a few hundred dollars in ads brought in about ten times that in new rides.
Study
The server behind a US business bank: card approvals, credit, and the APIs the phone apps later used.
Six months on a business bank’s servers. We built a service that approves or declines a card while the customer is still standing there, moved two generations of code onto one toolchain, and kept shipping credit, payments, and cash-flow tools on that same server.
Tell us the workflow, how often it has to run, and what it is allowed to see. We will price the cloud plan against the hardware, including the plan you would need once the cheap quota runs out.