Lola Vision Systems Bets on Software Licensing and Embedded AI

⚡

Key Takeaways

1Lola Vision Systems claims its first client and a dozen letters of interest for its future chips

2The company will now offer its licensed software on existing hardware

3Over 1 million dollars have been raised, and the startup has been selected for Battlefield 200

💡Why it matters — Lola Vision Systems aims to accelerate the adoption of embedded AI by simplifying the configuration of chip models and generating revenue before the release of its own semiconductors.

The Washington-based startup designs a compilation chain and chips to run AI models on-device. It claims to have a first client and a dozen letters of interest, and says it has raised just over $1 million. Meanwhile, it plans to license its software on existing hardware right away and has been selected for the Battlefield 200.

Letters of Interest, a First Client, and Over $1 Million

Lola Vision Systems reports having a dozen letters signed by corporate clients interested in its future chips and claims to already have one signed client. Tayo Adesanya states that the company has raised just over $1 million at this stage. To generate revenue sooner, he announces that the software will now be licensed on existing hardware, rather than waiting for the availability of in-house chips. The company has also formed a partnership with SCALE to collaborate with more microelectronics laboratories.

A Compilation Chain to Translate Models and Code into Instructions

The flagship product is software that translates AI models into executable instructions for a given chip, referred to as a "compilation toolchain" by Tayo Adesanya. The client provides their code and model, whether customized or open source, and the platform generates the instructions tailored to their chip. The company claims to have rebuilt this compilation layer and is developing its own semiconductors to further automate the process.

Accelerating Deployment Without Sacrificing Reliability and Consumption

Tayo Adesanya estimates that manual configuration of a model on new hardware takes about 200 hours before even testing begins. He argues that saving this time would allow critical sectors like aerospace to run more accurate models on their own data with reduced energy consumption. He emphasizes that precision and reliability are crucial for passing regulatory examinations and operational performance.

The Limitations of Current Approaches According to the Founder

According to Tayo Adesanya, many teams start with NVIDIA Jetson or open-source models, which he claims often turn out to be unstable or underperforming initially, requiring days or weeks of adjustments and then debugging. He asserts that even afterward, consumption frequently exceeds edge budgets or that the board does not deliver enough computation for medium to large models, leading to delays in recognition of their targets or misinterpretations. Washington, D.C.-based Lola Vision Systems positions itself among the startups offering an alternative to NVIDIA technologies for on-device AI.

Visibility at the Battlefield 200 and Project Trajectory

Lola Vision Systems has been selected for the TechCrunch Battlefield 200, which brings together 200 startups. Tayo Adesanya explains that he applied after about a year of development and signing a first client, aiming to present the company to a broader audience. He mentions his long-standing connection to TechCrunch from his time studying at Purdue and hopes to use the event to network, learn, and, he hopes, meet investors. The founder launched Lola Vision Systems in 2024 after a journey that began nearly 12 years ago with major hardware manufacturers, an experience that provided him with early market insights and, he says, motivated his entrepreneurial bet.