RealityEngines.AI, the machine learning startup co-founded by the former AWS and Google executive Bindu Reddy, today announced that it is renaming Abacus.AI and is launching its autonomous AI service in general availability.
In addition, the company also announced that it had raised a $ 13 million Serie A round led by Index Ventures Mike Volpi, who will also join the company's board of directors. Seed investors Eric Schmidt, Jerry Yang and Ram Shriram also participated in this over-subscription round, with Shriram joining the company's board of directors. New investors include Mariam Naficy, Erica Shultz, Neha Narkhede, Xuezhao Lan and Jeannette Furstenberg.
This new round brings the company's total funding to $ 18.25 million.
At its core, RealityEngines.AI of Abacus.AI's mission is to help companies implement modern deep learning systems in their customer experience and business processes without having to do the hard work of learning how to train the models themselves. In contrast, Abacus takes care of the data pipelines and the training model for them.
The company has worked with 1,200 beta testers and in recent months the team has focused primarily on helping companies build their own models, but also putting them into production. Current Abacus.AI customers include 1-800-Flowers, Flex, DailyLook and Prodege.
"My guess would be that of the one hundred projects launched in ML, one percent are successful thanks to so many moving parts," Reddy told me. "You have to build the model, then you have to test it in production - and then you have to build data pipelines and insert training pipelines. So even in the past few weeks, we've added a lot of features to allow these things to go into production more smoothly - and we keep adding on. "
In recent months, the team has also added new unsupervised learning tools to its range of pre-built solutions to help users build systems for anomaly detection on transaction fraud and account acquisition, for example.
Today the company has also released new tools for the debiasing dataset that can be used on already trained algorithms. The automatic creation of training sets, even with relatively small data sets, is one of the areas on which the Abacus team has focused for some time and now uses some of these same techniques to address this problem. In his experiments, the company's facial recognition algorithm was able to dramatically improve its ability to detect whether a black celebrity was smiling or not, for example, even if the training data set had 22 times whiter.
With today's launch, Abacus is also launching a new section on its website to showcase models from its community. "You can go build a model, edit your model if you wish, use your data sets and then you can actually share the model with the community," explained Reddy and noted that this is now possible thanks to the new price of Abacus model. The company decided to charge customers only when they put models into production.
The next important element on the Abacus roadmap is to create multiple connectors to third party systems so that users can easily import data from Salesforce and Segment, for example. In addition, Reddy notes that the team will develop more predefined solutions, including more work on language understanding and vision use cases.
To do this, Abacus has already hired more researchers to work on some of his fundamental research projects, something that Reddy says his funders are comfortable enough and more engineers to put that work into practice. He expects the team to move from 22 employees today to around 35 later this year.
