We’re excited to announce that Aquarium has raised $2.6M in seed funding led by Sequoia with participation from Y Combinator. As part of our funding, we’re partnering with Mike Vernal from Sequoia. The funds will be used to grow our team, continue to improve our product, and expand the business.
Aquarium is a machine learning (ML) data management system. A ML model is only as good as the data it’s trained on: Aquarium makes it easy to find labeling errors and model failures, then helps you curate your dataset to fix these problems and optimize your model performance.
In the year since we founded Aquarium, we graduated from Y Combinator, crunched millions of datapoints, and helped a wide set of customers improve their model performance by up to 18% in a single iteration cycle.
Quinn and I started Aquarium because we saw many amazing applications of artificial intelligence emerge across sectors as disparate as agriculture, manufacturing, logistics, medicine, and more. Simultaneously, we saw that most projects were still in the prototype stage and had trouble getting their models to adapt to the complexities of real world deployments. Although AI is going to fundamentally change the way that people do work, it hasn’t yet lived up to its potential.
In our previous experience as early employees at Cruise, we developed the machine learning models for self-driving cars and grappled with making these models perform at a sufficiently high performance level to be safer than a human. Now we are taking our learnings and are building world-class products that will enable the next wave of transformative AI applications in fields beyond self driving. We work with a variety of great customers that are deploying machine-learning models into production: UiPath for robotic process automation, AMP Robotics for recycling robots, and Sterblue for infrastructure inspection drones, to name a few.
We plan to use this funding to help our customers ship better models faster. We are investing heavily in making it easier to identify problems in model performance, find the right data to label to fix these problems, and retrain a measurably better model that’s production-ready. Our end goal is to make it easy for non-ML experts to build a model and continually improve it over time.
This is just the beginning of our journey, and we’d love to talk to anyone who would benefit from using Aquarium to supercharge their own ML pipelines. If you’d like to try Aquarium out for yourself, check out our website!