Why We Are Excited To Partner With Groundlight AI

2 min readApr 19, 2023


The manufacturing sector is a core engine of economic growth in the US, representing more than $2.35T in spend and greater than 10% of US GDP. It’s estimated that each $1 invested in the manufacturing sector added $2.79 of value to the broader US economy. Because of this, industrial automation and the next generation manufacturing facility has been discussed for over a decade. Despite steady progress, the pace of innovation has been slower than promised.

Groundlight AI aims to provide a needed step-change in industrial automation by democratizing computer vision integration. Groundlight’s platform empowers any developer, regardless of data science expertise, to quickly build robust vision solutions in minutes. The product collapses the computer vision deployment workflow into a single API that creates application specific models that are continually updated on the customer’s behalf.

Groundlight is led by pioneering technologists, Leo Dirac and Avi Geiger. Leo was a Senior Principal Engineer at Amazon where he led numerous deep learning and applied science projects, including the AutoML effort for Amazon SageMaker, and importantly, the integration between ML and robotic systems. Avi was a Principal Engineer and Architect at Microsoft and has been leading hardware products from design through the manufacturing process for over two decades.

We believe Groundlight is capitalizing on three major trends to drive a step-change in manufacturing automation:

The decreasing price of sensors has democratized IoT accessibility, unlocking the long tail of manufacturer demand. Groundlight is well positioned to grow with this newly enabled customer base. Today, customers can deploy computer vision with Groundlight without the need to have data in hand. As customers deploy IoT and access more data, they are equipped to deploy vision in new ways, faster. Going forward, applying sensor functionality to sensorless systems is less fanciful with Groundlight.

Infrastructure tooling has modularized the computer vision tech stack and continual, real-time analysis as a service is possible. SMB manufacturers typically rely on system integrators as consultants to surface insights about their facility. Modern infrastructure technology, the data tooling ecosystem, and proven deep learning techniques are fundamentally needed to match the value of integrators. Groundlight’s ability to pair these elements with real human judgement increases model performance.

Foundation models lower the cost of bootstrapping datasets needed to unlock vision capability for manufacturers. Labeling visual data is incredibly resource-intensive and manual. Techniques like zero shot learning are shifting the cost structure of developing the balanced dataset needed to train supervised learning models. Further, allowing non-technical personnel in a industrial setting to query the underlying model with natural language will change manufacturers’ dependency on third party integration services.

We are excited to partner with an experienced team in Leo Dirac and Avi Geiger, along with co-investors Madrona and Founders Co-op to help change U.S. manufacturing.

Our enterprise team continues to investigate how technology can create value in the industrial automation space. We are excited to partner with technical teams that are innovating in this area. If you are building a technical product in the industrial automation space, we would love to meet you.




A seed-to-growth venture capital firm that partners with exceptional entrepreneurs to build the world’s most transformative companies.