We’re on a mission to democratize state-of-the-art machine learning
The primary obstacle to effectively modeling complex, real world problems isn’t modeling itself — it’s data. Model performance is always limited by how well your data represents what you’re trying to predict. We can raise that limit using any model and in any domain using a foundational new embedding model.
Instead of relying on resource intensive deep learning to compensate for sparse, limited, high dimensional datasets, we can achieve better results by learning to create richer representations of your data before modeling. In the process, we capture complex, non-linear relationships in simple, statistical terms your model can understand.
Introducing Dark Matter, an embedding API with a unique objective function
We’ve developed a net-new machine learning algorithm that accounts for hidden relationships in your data.
By taking snapshots of the loss landscape and encoding them into embeddings that represent new dimensions in your data, we make these relationships clear for your model where they were previously considered noise.
The result is improved predictive accuracy of any model in any domain.
Feature Enhancement: a new step in the pipeline
Derived from a new statistical theory not present in the current literature, Dark Matter offers a faster path to superior model performance.
This technology constitutes a fundamental, new step in the data science pipeline that we call Feature Enhancement. It slots seamlessly in after feature engineering to learn how to create statistically optimal embeddings to train your model.
Feature Enhancement is just the beginning. This new technology will enable a range of downstream modeling capabilities that are not possible today, providing a platform for future innovation across the ML pipeline.
Our Vision
Ensemble aims to level the ML playing field, offering machine learning practitioners and researchers access to sophisticated modeling capabilities. We’re dedicated to rigorous scientific innovation that creates products that enable engineers and researchers to do more — augmenting their capabilities and intelligence.
Our Values
Stay Hungry
We never stop learning. We eagerly pursue innovation, owning our mistakes along the way. We set goals with a bias to action. We are unafraid and unjudging of failure, in ourselves or others. We achieve.
Stay Smart
We are flexible in thought and agile in our actions. We iterate quickly, and adapt readily to change. We put our team’s goals above our own. We don’t live in the past, worry about the future, or indulge in wishful thinking. We stay present, centering our attention to the task at hand.
Stay Humble
We value emotional intelligence. We behave and speak with kindness and respect, knowing our actions impact others. We are mindful that our view is just one of many. And strive to see situations with clarity, free of bias or emotion.
Team
We are scientists and engineers working together to push the boundaries of machine learning.
Advisors
Research
Feature Enhancement: A New Approach for Representation Learning (Whitepaper)
Discover a novel approach to representing complex, non-linear relationships inherent in real-world data.
Feature Programming for Multivariate Time Series Prediction (ICML)
Learn about a new framework for automated feature engineering from noisy time series data.Resources
Blog
Op-eds and thoughts on the state of machine learning and AIDocumentation
Developer support, API docs, quick-start guidePublished Research
Ensemble research, papers, and conferencesBacked by: