Galileo emerges from stealth to streamline AI model development – TechCrunch

As the use of AI becomes more common across the enterprise, the demand for products that make it easier to inspecting, discovering and resolving critical AI errors is increasing. After all, AI is precious — Gartner predicted in 2021 that a third of technology providers would invest $1 million or more in AI by 2023 — and debugging a flawed algorithm threatens to push development budgets up. A separate Gartner report found that only 53% of projects make it from prototype to production, presumably due in part to errors — a significant loss, if one were to add up the expenditure.

Vikram Chatterji, Atindriyo Sanyal and Yash Sheth, fed up with the high failure rate — and the fact that small (if important) data prep tasks, such as loading and cleaning data, still take up most of data scientists’ time — have Galileo, a service designed to act as a collaborative system for developing AI models. Galileo monitors AI development processes and uses statistical algorithms to detect potential points of system failure.

“There were no purpose-built machine learning data tools on the market, so [we] started Galileo to build the machine learning data tooling stack, starting with a [specialization in] unstructured data,” Chatterji told TechCrunch via email. †[The service] helps machine learning teams improve their data sets… by unearthing critical cohorts of data that may be underrepresented or inaccurate, while providing an all-round solution to encourage data scientists to proactively monitor data changes in production and identify errors and gaps in their models to reduce leaks in the real world.”

Chatterji has a background in data science and spent three years at Google as a product manager on the Android team. Sanyal was a senior software engineer at Apple, focusing primarily on Siri-related products before becoming an engineering lead for Uber’s AI team. As for Sheth, he also worked at Google as a staff software engineer and managed the Google Speech Recognizer platform.

With Galileo emerging today from stealth with $5.1 million in seed funding, Chatterji, Sanyal and Sheth set out to create a product that could scale across the entire AI workflow — from pre-development to post-production — as well as data modalities. such as text, speech and vision. Galileo is available in private beta and built to be deployed in an on-premises environment. The goal is to systematize pipelines between teams using “auto-loggers” and algorithms that highlight system-breaking issues.

Finding these issues is often a major pain point for data scientists. According to a recent survey (from the MLOps Community), 84.3% of data scientists and machine learning engineers say the time it takes to find and diagnose problems with a model is a problem for their teams, while more than one in four (26.2%) admit it takes a week or more to identify and resolve issues.

“The discussion about machine learning within the enterprise has shifted from ‘What am I using this for?’ to ‘How can I make my machine learning workflows faster, better and cheaper?’” said Chatterji. †Galileo …enforces the necessary rigor and the proactive application of research-backed techniques at every step in producing machine learning models… [It] leads to an order of magnitude improvement in how teams handle the messy, mind-numbing task of improving their machine learning datasets.”

Galileo fits into the emerging practice of MLOps, which combines machine learning, DevOps and data engineering to deploy and maintain AI models in production environments. The market for MLOps services could reach $4 billion by 2025, according to one estimate, and includes startups such as Databricks, DataRobot, Algorithmia and established companies such as Google Cloud and Amazon Web Services.

While investor interest in MLOps is growing, cash doesn’t necessarily translate into success. Even today’s best MLOps platforms cannot solve every common problem associated with AI workflows, especially when business leaders are unable to quantify the return on investment of these initiatives. The MLOps Community Survey found that convincing stakeholders when a new model is better, for example, remains a problem for more than 80% of machine learning practitioners.

Chatterji points to Kaggle CEO Anthony Goldbloom’s investment in Galileo — Goldbloom led the seed round with The Factory — as a sign of the company’s differentiation. Chatterji says Galileo currently has “dozens” of paying customers ranging from Fortune 500 companies to startup startups — revenue Galileo plans to use to triple the size of its 14-person team by the end of the year.

“Galileo has focused on turning around the otherwise arduous task of machine learning data inspection, to make it easy and provide intelligent data insights quickly,” Chatterji said. “The user only needs to add a few lines of code.”

To date, Galileo has raised a total of $5.5 million in venture capital.

Leave a Comment