Ensemble

Thanks Lux Capital for hosting the Lux AI Summit! We had a great time reconnecting with familiar faces and enjoyed the panels and demos :) . I was particularly excited about what they shared in “Embeddings are predictive of model uncertainty”, mentioned during the panel with Danqi Chen at Princeton, He He at NYU, Laurens van der Maaten at Meta, and Sasha Rush at Cornell Tech. This insight was significant. It highlights the fundamental importance of data representation and its impact on model development. Embeddings that effectively represent the underlying true behavior of the data will lead to the development of more certain models (and don’t require nearly as much complexity). Thanks again to Grace Isford for putting this all together.

For how much time machine learning engineers spend dealing with “data quality” issues, the term itself is often poorly defined. Until recently, the topic has been a relatively small part of the discourse in ML (despite its outsized impact on building models that work). In light of this, we have centered on a novel approach to thinking about data quality — plus how we can address it more fundamentally than before. Check out how this change in perspective led to groundbreaking improvements in machine learning and what we think should be a new step in the ML pipeline: https://lnkd.in/gZA-UeZt

Thrilled to be backing Alex Reneau, Zach A., and the rest of the Ensemble team???? We knew quickly after meeting Alex & Zach that they were a team we wanted to work with. Not only do they have deep domain expertise, but they are also opinated, focused and humble - our favorite kind of founders ?? We're excited to share publicly what they've been working on over the past year with Dark Matter, a net-new representation learning algorithm that allows organizations to do more with less data. Ensemble’s proprietary technology and approach to data viability allow enterprises to solve old problems better and new problems for the first time. This is just the beginning for Ensemble, and we look forward to being on this journey with our friends at M13, Motivate Venture Capital & Amplo. Jason Spinell Salesforce Ventures Read more about our investment here: https://lnkd.in/gJvVpmRV https://lnkd.in/gAFh5yTq

It is clear that one of the most important ingredients in building successful ML models is data quality. Ensemble is reimagining the data science pipeline with an approach that enriches data and improves model performance all while reducing compute needs. We believe Ensemble’s product will allow new breakthroughs for both data scientists and enterprises across industries. Their approach to implementing AI at scale will help companies overcome previous barriers like data quality issues or resource constraints. We are so excited to support Alex Reneau and Zach A. on this mission, and we believe Ensemble will one day be utilized in every ML pipeline. Honored to have met this team at the start of their journey because the future is bright! cc: Salesforce Ventures, Motivate Venture Capital, Amplo https://lnkd.in/dmAe-kKV

We are excited to share that we have raised $3.3m in a seed funding round led by?Salesforce Ventures?with participation from?M13,?Motivate Venture Capital, and?Amplo. We're also excited to introduce Dark Matter, our embedding API which learns the optimal way to approximate relationships in data for any machine learning task — getting ML models working faster and better without the need for extensive feature engineering. This new foundational approach frees up data scientists to focus on experimentation, improving their ML pipelines, and working on problems that were previously unable to be modeled and are beyond thrilled that Salesforce shares our vision. https://lnkd.in/geRTqtYR For those looking to create more robust ML models in less time and headache, please reach out - we'd love to hear from you:?https://lnkd.in/gVMfVwCB

In the current push to advance AI for all aspects of industry, a critical shift in focus is needed: from model complexity to the quality of data. The current fascination over modeling solutions (but especially LLMs) overlooks a key element - the data on which these models rely. High-quality, specific datasets are not just a nice-to-have; they're the foundation of effective AI & Machine Learning. The mantra ‘bigger is better’ in model development is often misguided. In fields like finance and healthcare, the field is upper-bounded in performance with current models trying to account for every possible factor. The solution? Smaller, specialized models honed with meticulously selected data. But mere data curation isn’t enough. The future of AI advancement hinges on our ability to enhance and refine our data. As we look ahead, the real innovations in AI will be in how we enrich and refine the data that powers our models, not just in the models themselves.