Masterful AI: Building Better Machine Learning Models

November 17, 2021

There’s often a magic moment when we hear great pitches from founders - when their vision clicks with us and leads us to say “of course, that’s exactly the way things should work”. That was precisely our response when the Masterful AI team shared their insights and vision for a smarter, faster way to build more accurate machine learning models. We’d seen the challenges of building machine learning models when we were operators, and much more recently in working with many of our SaaS and fintech companies. And as ML models make more and more decisions that impact our lives, it’s increasingly important to make sure those models are unbiased and accurate. It was an easy decision to lead Masterful’s seed financing and to add to our portfolio of developer-focused product companies.

WHO: Tom Rikert, Sam Wookey and Yaoshiang Ho have AI research roots at MIT, Stanford, and Google, and have shipped products to millions of users at Microsoft and YouTube. As a group, they’ve dealt with the frustration of trying to develop production-ready ML models. Those practical experiences, combined with their extensive academic backgrounds, prepared them to build Masterful.

WHAT: Masterful is creating a smarter, more automated way to build ML models. The AutoML platform they announced recently reduces the data labeling required and shortens the time needed to achieve a performant model. Masterful can be integrated with just a few lines of code via a simple API. On average, Masterful can reduce model error by at least 40% and shorten deployment time by up to 50%.

HOW: Masterful helps build better models by using unlabeled, augmented, and synthetic data to improve models and then automatically test and tune model training loops. It all starts with tackling the most expensive and inefficient part of model development - getting accurately labeled training data. While some labeled data is necessary, Masterful’s reduces the amount of labeled data required to train a model by 10x. It combines small amounts of labeled data with unlabeled, augmented, and synthetic data that can be managed by code instead of people. Then Masterful eliminates manual experimentation for testing different datasets and training algorithms. Instead, Masterful automatically tests and evaluates a suite of training algorithms and hyperparameters, ultimately producing an optimal training policy and a visual explanation of what was chosen and why. All of this is possible because of the deep learning pipeline at the core of Masterful’s platform.

WHY: The work that Masterful is doing is critical because it modernizes the primitive way in which ML models get built. We’re at a moment in time when machines are going to start making ever more important decisions for us. And those decisions will impact countless lives. Masterful eliminates the time delays, costs, and potential for bias inherent in human data labeling. With more accurate ML models, we can feel better about the decisions being made without human input and avoid the catastrophic mistakes that can result from less sophisticated model building approaches.

If you’re looking to decrease labeling costs and increase model accuracy, you can get a free report here showing what Masterful can do for your model. And if Masterful’s mission is compelling to you, the company is hiring.