Introducing 🚀 – our free batch evaluation tool for Large Language Models. Get started for free!

Sematic raises $3M to build an open-source Continuous Machine Learning platform

November 17, 2022
Emmanuel Turlay
Emmanuel Turlay
Founder, CEO
Sematic raises $3M to build an open-source Continuous Machine Learning platform

Today, we are announcing a $3M seed funding round led by Race Capital, and accompanied by Y Combinator, Soma Capital, Leonis Capital, Fundament, and Pioneer Fund.

Thanks to this amazing set of backers, we built a world-class engineering team to deliver on our mission to provide Machine Learning teams across the industry with the easiest way to prototype, automate, and productionize end-to-end ML pipelines while getting unprecedented visibility and observability out-of-the-box.

Machine Learning development needs a breakout moment

A decade ago, tools such as Ruby on Rails and Heroku enabled generations of web developers of all skills to build prototypes into unicorns.

Suddenly, the web development community aligned around best practices, standards, and tools. Productivity exploded enabling countless modern companies to build and ship products faster and safer.

This shift has not yet happened in the Machine Learning space. We want to make it happen.

The vast majority of ML projects never make it to production

Today, the vast majority of ML projects never make it to production due to their lack of testability, usability, reproducibilty, automation, and scalability. Machine Learning has the potential to greatly optimize productivity, revenue, safety, and resource usage across all industries but there remains a vast gap between its theoretical potential and its practical applications.

As we have witnessed throughout our careers across academia and the industry, the main reason for this vast gap between aspirations and applications is a mismatch between the available tooling and the expectations of the workforce.

Current tools are simply too difficult to use and do not provide the necessary guarantees to power robust production systems developed by growing teams.

Learnings from the robotaxi industry

Tudor, Emmanuel, and Josh

The Sematic founding team spent the last four years building ML Infrastructure for Cruise, the first robotaxi company to offer a commercial service.

At Cruise, we successfully bridged the gap between the complexity of large scale cloud infrastructure and the expectations of a growing ML workforce.

We empowered hundreds of ML engineers and scientists to build, deploy, operate, and debug arbitrarily complex end-to-end training pipelines by themselves, without having to hand over projects to infrastructure teams.

We believe these learnings should transfer to the rest of the industry.

Continuous Machine Learning for safer, better models

ML models are not one-and-done projects. The underlying training data grows and changes constantly. Pandemics, macroeconomic conditions, usage patterns, trends, world events, etc., data evolves constantly.

Without the ability to build automated feedback loops that continuously collect fresh data to retrain models, learn from mistakes, and from emerging patterns, models are bound to remain toy projects at best, losing relevance over time and turning into liabilities.

These feedback loops have to come with strict guarantees around traceability, reproducibility, and observability to ensure safety, compliance, and scalability. Sematic aims to build the open-source standard for Continuous Machine Learning.

With the easiest onboarding and lowest barrier to entry, Sematic aims to attract ML engineers and scientists of all skills, and empower them to build continuous learning end-to-end pipelines, and scale them from laptop prototypes to fully automated cloud systems.

Stay tuned, we are just getting started

We are incredibly grateful to have gathered the trust of stellar investors, and a team of world-class engineers, as we aim to revolutionize ML development, and bring it the break-out moment it needs for ML to deliver on its promise.

We are just getting started.

Subscribe to our mailing list

Receive release notes, updates, tips and news straight into your inbox.
Average frequency: every couple of weeks. No spam ever.

Subscribed, thank you!
Oops! Something went wrong while submitting the form.