Former chip consultant Tayo Adesanya launched Lola Vision Systems in 2024 to remove a bottleneck in edge AI: manually preparing a model for new hardware can take about 200 hours before testing even starts. The Washington, D. C. company has built a compiler toolchain that translates client code and models into chip instructions, and it is developing its own semiconductors. The approach matters for business because faster deployment lets mission-critical operators run more accurate models on their own data at lower power.
Compiler toolchain and early customer traction
Adesanya spent almost 12 years helping large manufacturers select microchips and AI processors, work that shaped his view of computing demand. Lola Vision Systems positions itself as an AI infrastructure supplier for on-device inference, competing with alternatives to NVIDIA technology. The company reports a dozen corporate customers with letters expressing interest in its future chips, plus one signed customer already. Total funding to date stands at just over $1 million.
The core product accepts a client code base and a chosen AI model, whether custom-built or open source, and converts both into instructions a specific chip can execute. Adesanya describes the compiler layer as rebuilt to automate steps that teams now perform by hand. The company will license this software for existing hardware to generate revenue before its own chips become available. It has also partnered with SCALE, a microelectronics workforce development program, to work with more semiconductor labs.
The background is the common starting point for many device makers: NVIDIA Jetson compact modules or open-source models. According to Adesanya, those combinations often break or perform poorly out of the box, forcing teams to spend days or weeks on basic bring-up and additional weeks on debugging. Power consumption then exceeds edge budgets, or the board lacks compute for medium to large models. The result is recognition systems that miss targets or misread objects in cameras, drones and similar equipment.
What faster on-device deployment means for business
For aerospace and other regulated operators, shorter setup changes project economics because engineers can test larger and more accurate models on proprietary data without moving work to a data center. Lower power use makes continuous operation on cameras and drones more practical and reduces redesign around thermal and battery limits. Small firms gain a way to reach a working prototype without a large porting team, while large manufacturers can standardize deployment across several chip types.
The limits deserve attention during procurement. Lola Vision own chips are not yet available, so near-term buyers are evaluating software running on third-party silicon rather than a complete hardware stack. Accuracy, reliability and power figures still need validation on the customer model and board, since edge performance depends on sensors, data and operating conditions. Buyers should ask which chips and model architectures are supported, what testing is included, and how updates and regulatory documentation are handled.
A practical marker is the company appearance at TechCrunch Disrupt in San Francisco on October 13-15 as part of Battlefield 200. Conversion of the dozen interest letters into paid software licenses, followed by paid pilots on existing hardware, would confirm demand for the toolchain alone. Adesanya said he expects investor meetings at the event, so funding announcements and named deployments are the next signals to track.
