Neptune.ai raises 32 mln PLN in Series A for its “Github for machine learning ” platform

The Series A funding will be used to hire engineers across Europe, allow Neptune.ai to build on its recent 4X growth, and expand its customer base of more than 20,000 machine learning engineers and 100 commercial customers from Roche, The New Yorker and InstaDeep.

13.04.2022

Neptune.ai – the “Github for machine learning” platform helping engineers train machine learning models faster and more efficiently – has raised $8M in Series A funding. The round was led by global VC firm Almaz Capital, with participation from previous investors btov Partners, Rheingau Founders, and TDJ Pitango Ventures. 
Initially created by machine learning engineers from deepsense.ai – a group led by former Facebook and Google alumn Piotr Niedźwiedź – the platform was spun out in 2017 to help teams across the globe store and manage a vast amount of metadata created by machine learning models. 
This metadata, which includes references and insights from the datasets being used, code versions, environment changes, hardware, evaluation and testing metrics, predictions and much more, is ever-evolving. Each time a model goes through a different iteration, taking a different training path, or using different datasets, it creates a complex web of version histories. This can be difficult to navigate when things are running smoothly, yet when a machine learning model hits a snag and fails, it can prove both costly and timely to unpick what went wrong and when. 
Github for data science
With Neptune, anyone working with such machine learning models can replace and augment the folder structures, unwieldy spreadsheets, and naming conventions common among today’s teams with a single source of truth. 
It not only manages the infrastructure needed to log and store this metadata, but it provides a central place for teams to view, organise, share, compare, query and collaborate on all metadata generated during an entire machine learning lifecycle. It helps companies get the most out of their resources by keeping a record of all the paths they have taken and it automatically creates versions to backtrack failed runs. These help engineers establish which dataset and parameters were used, and help them to recover quickly. 
The entire platform offers its customers unprecedented insight into the evolution of their models, while freeing them up to focus on the products they’re building. Neptune.ai also saves companies time and money by automating the metadata bookkeeping and ensuring that companies don’t have to hire extra people to implement loggers, maintain databases or teach people how to use them. 
Just as Github transformed how software engineers log, store, manage and share their code, Neptune.ai is on a mission to become the Github for machine learning. It wants to become the foundation tool of any end-to-end ML platform and one which enables machine learning teams to have the same level of control and confidence when developing and deploying models as software developers have when writing and developing apps. 
Transforming machine learning 
Today, the Neptune.ai platform helps more than 20,000 data scientists, engineers and businesses make more informed machine learning decisions. This includes 100 commercial customers including Roche, The New Yorker, NNAISENSE, and InstaDeep. Its platform has grown four-fold over the past year and this saw Neptune.ai listed as one of CB Insights’ Top 100 AI startups to watch over . 
The Series A funding round takes the total raised by the Polish startup close to  $13 million. This latest investment will enable Neptune.ai to deliver an even better experience for developers by rapidly expanding their product and engineering teams across Europe. Neptune.ai also plans to grow its customer base. As machine learning standardises and becomes more common, the aim is for Neptune.ai to become the most-used tool for storing and organising model metadata globally.
Piotr Niedzwiedz, co-founder of Neptune.ai, said: “When I came to machine learning from software engineering, I was surprised by the messy experimentation practices, lack of control over model building, and a missing ecosystem of tools to help people deliver models confidently. So when ML engineers at my previous company showed me a tool they built for experiment tracking, I knew it had massive potential. Fast forward to today, and we are one of the most popular tools for experiment tracking and model registry on the market. Thanks to the backing of Almaz Capital and our other investors we will continue building a better product that machine-learning engineers and data scientists can use far into the future.” 
Pavel Bogdanov, general partner at lead investor Almaz Capital, said: “As more companies adopt machine learning, the demand for tools that help operationalise and control model development and deployment grew substantially in 2021. Yet the world is still at the infancy of machine learning adoption, and we expect the MLOps market only to grow from here. What we liked about the Neptune.ai team was the clear vision to create the best-in-class, foundational component of the MLOps tool stack instead of trying to solve this problem end to end. With a fast-growing customer base and a focus on providing the best developer experience, we believe they can become the go-to solution for model metadata management for machine learning teams everywhere.”    
Daniel Star, managing partner TDJ Pitango Ventures, said: “The MLOps tools market is rapidly evolving from individual experiments to group collaboration. Neptune today provides the best toolkit for machine learning development teams to collaborate and achieve better outcomes. We have been with Neptune since 2018 and I am proud of what this team has achieved. We strongly believe in further scaling of the business, especially with such a dynamically growing global market.”
Ronert Obst, Head of Data Science at international retailer, the New Yorker, said: “What we like about Neptune is that it easily hooks into multiple frameworks. Keeping track of machine learning experiments systematically over time and visualizing the output adds a lot of value for us.”