Our Recent Posts


  • anushakolagani9

Bringing AI to Life with Unstructured Data: Zeff Has Joined DataRobot

We founded Zeff.ai with a mission to bring automated machine learning to teams working on the naturally complex datasets that arise from building software products and services, including datasets with images, audio, and unstructured data. Initially, we were just excited to see applied deep learning helping a variety of businesses. That excitement grew into an obsession to help as many teams as possible use AI to improve what they are building.  As a company on a mission to power next-generation AI applications with unstructured data, we have long admired DataRobot and its ability to reimagine how businesses can continuously unlock new opportunities with diverse types of data. Earlier this year, for example, DataRobot enhanced their Automated Machine Learning product and rolled out Visual AI, which gives you the ability to easily incorporate image data into your machine learning models alongside tabular and text-based data types. With Visual AI, DataRobot has shifted the market for image-based machine learning and computer vision.  As DataRobot strategically expanded its vision to encompass the entire AI journey from idea, to data, to business value,  it became increasingly clear that Zeff.ai’s mission was not only synergistic to that of DataRobot, but that we could reach new heights together.  Try the world’s leading enterprise AI platform now This innovative thinking paired with the ability to push the boundaries of what’s possible with AI are the main drivers for why Zeff decided to join forces with DataRobot three months ago. By acquiring the company that we founded in 2017, DataRobot is beefing up its abilities to enable its customers to use unstructured data, like video and audio to power their AI applications.  The applications for AI that involve unstructured data are practically endless. Here are a few we’re particularly excited about tackling:  Risk Assessment  Most risk models — especially for financial risk — have been built using structured data. However, when unstructured components like images are involved, models require a human touchpoint. This leads to a bottleneck with underwriters and risk experts who must manually review images tied to each assessment. Now thanks to the invention of deep-learning and computer vision, all of the unstructured data can be considered and incorporated.  Property Value Evaluation House valuations or prices are often manually evaluated by an appraiser who is unable to accurately value your landscaping or the style of your kitchen (think granite countertops or stainless steel). Or for online real estate sites such as Zillow, the valuations are determined by algorithms that are unable to place a valuation using the visual features in the photos. Using Visual AI, you can consider all of the data that matters, including images and text to create a much more comprehensive evaluation based on a wider set of data perspectives. Automation of Audio and Video Assessments To date, many activities, like remote proctoring, hiring, and manufacturing quality assurance have required humans to physically participate in the review process. This is a major bottleneck to productivity and growth. Additionally, humans can impact quality with irregular review processes, and can also introduce unintentional bias and opinion. Language Assessment Having native speakers listen and evaluate non-native speakers is a common use case. Now, AI can automate your video and audio assessments by listening and automatically scoring audio for the desired level of language competency.  Remote Education Measuring or evaluating a competency or behavior can be enabled with video or audio features. Historically, evaluating a child’s speech impediment required a human review, where these types of technologies will allow us to push forward into this beautiful future where more students will benefit from frequent prescriptive feedback.  Defect Detection Many manufacturing processes involve manual review for quality assurance. For example, looking for  paint scratches on vehicles or trying to detect abnormal visual queues on a silicon wafer to indicate misprocessing. In addition, there is also a lot of value in predictive preventative maintenance and using audio or other unstructured datasets to detect strain or abnormal behavior on manufactured parts, components and equipment.  It’s clear the opportunity for unstructured data is massive and the Zeff team is excited to be a part of DataRobot’s journey. We look forward to continuing to push the boundaries of what’s possible with AI for customers everywhere — and to everyone who has believed in us from our Angel investors to our families, we thank you.

  • LinkedIn
  • Facebook

©2018 by B-AIM