Openlayer, creator of tools for testing and validating machine learning models, raises $4.8M

Openlayer Inc., creator of a new platform for testing artificial intelligence and machine learning models, announced today that it has closed a $4.8 million seed funding round.

The round was led by Quiet Capital, with participation from Picus Capital, YCombinator, Hack VC, Liquid2 Ventures, Mantis VC, and a number of angel investors, including Instagram co-founder Mike Krieger and Instacart co-founder Max Mullen.

Openlayer was created by former Apple Inc. machine learning engineers Gabriel Bayomi, Rishab Ramanathan, and Vikas Nair. The startup explains that there are many existing software testing platforms available to developers, but most of them are designed for deterministic systems where certain inputs produce expected outputs. . Therefore, it is not suitable for probabilistic models. Until now, developers haven’t had a reliable way to test their models, the startup says.

“We faced the same problem over and over while working in 15 different teams at Apple. There is no standardized way to test and collaborate on machine learning models,” said Nair. I will explain. “This is why errors are often not discovered until after shipment, which can lead to disastrous consequences.”

Openlayer changes that, allowing teams to systematically improve their models and datasets before shipping. For example, it can help teams validate the consistency of training and validation datasets, revealing meaningful discrepancies between training, evaluation, and operational data. Confirm that your model meets your target performance benchmarks, and validate whether generating synthetic data and injecting noise is robust enough for edge use cases.

Other features of Openlayer include ensuring fairness of model behavior across data subpopulations and tracking different versions of models and datasets to compare performance. It can also explain the behavior of the model by revealing which aspects of the dataset were used for prediction.

Generative AI bots like ChatGPT have created all sorts of buzz and excitement lately, but more tools are urgently needed in terms of testing, security and validation of AI, says vice president of Constellation Research Inc. said Andy Thurai, Principal Analyst. He explained that model testing and validation is an important part of the machine learning operational process, as visibility is a key consideration for companies deploying AI solutions.

“While most MLOps platform providers build such capabilities as part of their overall experience, Openlayer has a particular focus on validation components to help fast-moving companies, with several It gives us an advantage,” said Thurai. “Openlayer’s validation options for verifying the integrity of training and validation datasets, and its ability to reveal discrepancies between training, evaluation, and production data, are interesting and unique. Model performance measurements. , model tracking, dataset storage, and model explainability tools are all standard features found in most other MLOps platforms.”

According to Openlayer, many Fortune 500 companies use the platform to test and validate machine learning models, find mistakes, and diagnose why problems occur. Going forward, the startup plans to expand its team and improve the capabilities of its platform to handle additional machine learning tasks. We also want to add more sophisticated guardrails to allow users to test models as they iterate.

Bayomi said his vision is for Openlayer to be the go-to hub for any team building new models. “Achieving this requires building a development pipeline that provides strong insights into models and data at every stage, pre-deployment and post-deployment,” he said.

Image: Free Pick