Lola Vision Systems targets the 200 hours of setup between an AI model and a new chip

Many teams putting AI on a camera or drone start with an Nvidia Jetson module or an open-source model, according to Tayo Adesanya, who founded Lola Vision Systems in 2024 ^[12][5][1]. He said those setups "often break or run poorly out of the box, so teams spend days or weeks getting them to run at all, then even more weeks debugging until the models are usable." ^[12] The trouble continues after the model runs. "Even then," he said, "power consumption often blows edge computing budgets, or the board can't deliver enough compute for the medium to large models the product actually needs to run successfully." ^[13]

By Adesanya's account, teams spend setup hours before a single test result exists ^[11]. At a 40-hour week, his figure of roughly 200 hours comes to five weeks of one engineer's time on a new chip ^[10].

Lola Vision's pitch has two parts. One is a compiler toolchain the company says it has rebuilt, the software that translates a model into instructions a particular chip can run ^[2][3]. The other is its own semiconductor chips, now in development, according to the company ^[3]. The software is what a customer can pay for first. "To get revenue sooner, we will now license our software on existing hardware," Adesanya said ^[6].

That order matters to a team weighing a hardware switch. A compiler that can target many boards makes it cheaper to move between chip vendors. Lola Vision also intends to become one of them ^[3]. The account does not include a list of the existing chips the software supports.

For aerospace and other mission-critical buyers, Adesanya put accuracy next to the hours. "Speed is only part of it," he said, adding that faster setup gives those companies time to "run more accurate models on their own data, at a lower power." ^[16] "For these customers," he said, "accuracy and reliability aren't nice to have. They determine whether a product passes regulatory review and whether it works reliably in the field." ^[14]

A team can answer two questions about itself before talking to Lola: how often it moves a model onto a new chip, and whether on-device accuracy decides regulatory approval. A team that rarely changes chips and has slack on accuracy pays the setup cost once and can keep a young supplier out of its build. If chip changes are rare but the product faces review, a pilot is mostly a check that accuracy and power hold on the team's own data, with saved hours as a side benefit. Frequent chip changes with loose tolerances are the case the pitch fits, because the porting cost repeats with every board. The fourth case, frequent changes under regulatory review, is the strongest case for the toolchain and the deepest dependence on a startup. There, the pilot that settles it is the team's own model on its current board, measured in hours to a first valid test and in accuracy against the bench version.