Galileo, a platform for AI model development, today announced that it has raised $18 million in Series A funding led by Battery Ventures, with participation from The Factory, Walden Catalyst, FVV Ventures, Cagle founder Anthony Goldblum and other angel investors. . The new funding brings the company’s total to $23.1 million and will be used to grow Galileo’s engineering and go-to-market teams and expand its core platform to support new data technologies, CEO Vikram Chatterjee told TechCrunch in an email.
As the use of AI becomes commonplace across the enterprise, there is an increasing demand for products that make it easy to diagnose, detect, and fix critical AI errors. In a recent survey (from the Mloops community), 84.3% of data scientists and machine learning engineers said the time required to identify and diagnose model problems was a problem for their teams, with one in four (26.2%) admitting it. It will take a week or more for them to find and fix problems.
Some of these cases include mislabeled data where the labels used to train the AI system contain errors, such as an image of a tree labeled “houseplant.” Others are related to data drift or data inconsistency, which is when the data is generated to make the AI system less accurate (think of a stock market model trained on pre-pandemic data) or the data does not adequately represent the domain (for example, a data set of headshots with light skin rather than dark skin It has people who have).
The Galileo platform aims to organize AI development pipelines across teams using “auto-loggers” and algorithms that detect system-throwing issues. Deployable around campus, Galileo sCals in the AI workflow – from pre-development to post-production – as well as unstructured data formats such as text, speech and vision.
In data science, “unstructured” data usually refers to data that is not organized according to a defined data model or schema, such as receipts or sensor data. Atindrio Sanyal – Galileo’s second co-founder – explains that Excel and Python script-based processes to ensure quality data are fed into models are manual, error-prone and costly.
A screenshot of the Galileo Society Edition. Image Credits: Galileo
When users check their data with Galileo, they immediately recognize long-standing data errors such as mislabeled data and unrepresented languages. [and] Garbage data that you can take immediate action on in Galileo by removing, renaming, or adding more similar data from a product, Sanyal told TechCrunch in an email interview. “It was critical for teams that Galileo supports end-to-end machine learning data workflows – even when a model is in production, Galileo allows teams to automatically identify data slips and uncover high-value data for subsequent training.”
The co-founding team at Galileo has spent more than a decade building machine learning products, where they faced the challenges of developing AI systems directly, he said. Chatterjee led product management at Google AI, Sanyal led engineering at Uber’s AI division and was an early member of the Siri team at Apple. Galileo’s third co-founder, Yash Sheth, is another Google veteran, having previously led the company’s speech recognition platform team.
The Galileo platform is part of a growing software category known as MLOps, a set of tools for deploying and maintaining machine learning models in production. It is in high demand. According to one estimate, the MLOps market could reach $4 billion by 2025.
There’s no shortage of startups going after space, like Comet, which raised $50 million last November. Other vendors with VC backing include Arise, Tecton, Diveplan, Iterative and Taiwan-based InfuseAI.
But despite launching just a few months ago, Galileo is already paying customers from “high-growth” startups to Fortune 500 companies, Sanyal said. “Our customers are using Galileo when building machine learning applications such as hate speech recognition, identifying caller needs in contact centers and augmenting the customer experience with conversational AI,” he added.
Sanyal expects the launch of Galileo’s free offering — the Galileo Community Edition — will further boost subscriptions. Community Edition allows data scientists working on natural language processing to build machine learning models using some of the tools included in the paid version, he said.
“With Galileo Community Edition, anyone can register for free, add a few lines of code while training the model on labeled data or to quickly inspect, find and correct data errors, or train the model on labeled data to select the correct data. Give the next tag using the powerful Galileo UI,” he added.
Sanyal refused to share revenue figures when asked. But San Francisco-based Galileo’s headcount rose from 14 people in May to “more than 20” people, he said.
