Why We Invested in Quix.ai

Bringing Real-Time Machine Learning From Formula 1 Into Your Organization

By Leopold Larach, Early stage VC @Project A

There are very few industries in the world that are as data driven as Formula 1. In an average race car, 300 sensors are each sending 1500 data points per second, monitoring every aspect of the car from tyre conditions and break temperatures to airflow and vibrations in the wings. To make sense of roughly 30,000,000 data points per minute and support split-second decision-making, thousands of machine learning models are run in real-time to gain a competitive edge with endless options for different analyses. As an example teams might optimize their pit stop strategies, taking into account predicted lap times for each driver in the field, the likelihood of changing weather conditions and gradual tyre degradation. Sounds like an incredibly complex task and, yes, it absolutely is.

When we met founders Mike Rosam and Tomas Neubauer, we knew that their combination of entrepreneurial drive and deep technical expertise from the forefront of live data processing made them the perfect team to partner with. Having built and maintained the stream processing engine at McLaren Racing that supported on-track decision-making in real-time, we could really feel their frustration about the lack of a similar product offering in the market and the urgency to change this.

At Project A, we firmly believe that we’re still in the early innings of a data revolution: to be successful in increasingly competitive markets, companies will have to become highly data-driven organizations and leverage insights to create differentiated customer experiences and drive operational excellence. While historical insights from data analytics and science can inform strategic decisions, we think live applications will push the boundaries and unlock a whole new era of use cases beyond imagination (see for example operational analytics). Another recent example of hyper-personalisation is TikTok’s recommendation algorithm: it learns almost in real time about a user’s preferences in that very moment and seems to make much more accurate predictions than other social media giants, creating a uniquely tailored experience that is key to the entire platform.

On the infrastructure side, real-time machine learning at scale will require shifting from an architecture that handles data in batches to running on event streams, where ingestion, transformation and analysis is performed in-memory on each data point as it arrives (see here for a technical summary). Event streaming as a messaging paradigm has become increasingly popular due to the proliferation of microservices and the cloud. Confluent is leading its commercialization and is built on top of Apache Kafka, which was originally developed at LinkedIn.

However, companies wanting to do machine learning on event streams need to make significant upfront investments in additional infrastructure, even if they are already using events to orchestrate microservices. While large enterprises can afford to do some ‘undifferentiated heavy lifting’ internally, we believe in a future where no organization needs to take on this burden and developers can spend more time on the application layer and on creating customer experiences rather than on maintaining infrastructure.

Screenshot: Quix team during a Zoom meeting

After over a year of intense development, today Quix launches its platform to democratize data streaming from end to end. From ingesting, exploring and visualising event streams to building, testing, deploying and monitoring ML models on top and even triggering downstream actions — Quix has got you covered. To win this market, developer experience is key: the Quix platform is based on Python, the lingua franca for Data Scientists, and seamlessly plugs into engineering workflows and industry standard tooling, while most popular event streaming engines are based on Java/Scala and require significant additional stitching of open source technologies.

At Project A, we are incredibly excited to announce that we led their Seed round in autumn 2020, partnering with Passion Capital and a range of angels such as Frank Sagnier (CEO, Codemasters), Ian Hogarth (Co-founder, Songkick), Chris Schagen (CMO, Contentful) and Michael Schrezenmaier (COO, Pipedrive). We could not be more happy to partner with such a great team and to help them bring cutting edge technology from the Formula 1 race tracks to developers’ browsers and command lines!