2019 AI/ML Roundup
Top AI/ML vendor news for 2019.
New AutoML. Several established vendors added automation to their data science and machine learning product line.
- IBM launched AutoAI, a component of IBM Watson Studio.
- SAS added AutoML components to SAS Visual Data Mining.
- Google Cloud Platform launched Cloud AutoML Tables to beta.
- Alteryx announced the beta release of Assisted Analytics.
- Databricks announced an AutoML capability in August.
Improved AutoML. Existing players enhanced their offerings.
- H2O.ai added many new bits to Driverless AI, including:
- Bring Your Own Recipes
- Interpretability enhancements
- JDBC connectivity
- Deployment to Amazon Lambda
- Deployment to a local REST server
- R API
- Follow The Regularized Leader (FTRL)
- Amazon Web Services released Amazon Forecast to General Availability.
- AWS also announced six new components for Amazon SageMaker, including an AutoML component.
- Microsoft announced a drag-and-drop interface to Azure Machine Learning.
Out with the old, in with the new. SAP bought KXEN in 2013, integrated it with HANA and rebranded the product as SAP Predictive Analytics. That
dog software was getting long in the tooth, so SAP removed it from the catalog and added automated machine learning to SAP Data Intelligence. The latter move came too late for SAP to qualify for Gartner’s 2020 Magic Quadrant for Data Science and Machine Learning Platforms, so we won’t be able to laugh at SAP when the report comes out.
MLOps is hot. Everyone discovered MLOps last year. A slew of startups specializing in MLOps emerged, including Algorithmia, Datatron, Iguazio, and Seldon. Open Data Group, a Chicago-based consultancy founded in 2002, rebranded as ModelOp.
Several AI/ML players added capabilities:
- SAS announced SAS Open Model Manager.
- IBM added model drift detection to IBM Watson OpenScale.
- Microsoft Added MLOps to Azure Machine Learning.
- Databricks announced MLFlow 1.0 for MLOps.
H2O.ai delivered a product brief for MLOps but forgot to deliver a product.
Containers are hot. Everyone jumped on the containerized software bandwagon:
- HPE acquired BlueData and repackaged the product as on-premises PaaS for machine learning.
- IBM acquired Red Hat, then ported IBM Watson Studio and other software products to Red Hat OpenShift.
- Dataiku released Data Science Studio Version 6, with k8s support.
- Domino Data Lab released Domino 4.0 on Kubernetes.
I can’t wait to see what happens when the crowd decides Kubernetes is overhyped.
New kids on the block. Several new AutoML players emerged from stealth, including Aible, dotData, and SparkCognition. Each of these vendors offers some interesting features. All three are tiny. SparkCognition’s cybersecurity business drives its valuation and funding, AI/ML platform revenue is peanuts.
New funding. Several startups closed funding rounds.
- Databricks swallowed $650M in two rounds, a Series E and a Series F
- SparkCognition hooked a $100M Series C round
- H2O.ai landed a $75M Series D round
- Algorithmia secured a $25M Series B round
- dotData snared a $23M Series A round
- Datatron planted a $2.1M Seed Round
- Samsung Ventures invested an undisclosed amount in Iguazio
Dataiku didn’t get any new money but announced a bullshit valuation.
Upper management (CEO) creates an abusive work environment belittling employees, blaming them when he is caught in lies, and manipulating others to vilify individuals who upset him.
Big Squid just let go about half of its workforce and even cut entire departments. While I still like my coworkers, we kept a lot of people that, in my opinion, didn’t earn their spot. Work is a lot more stressful now, and I miss the greater team environment we used to have.
Ironically (for a predictive analytics company), one of the least data-driven companies I have worked for, not to mention, Big Squid has yet to build and implement its own first predictive model. Extreme lack of transparency perpetuated by the CEO, who also happens to be in a romantic relationship with a member of Big Squid’s customer success team. No visibility into the company’s customer retention or sales…
Other than that, things are going pretty well at Big Squid.
Much Ado About Nothing. Cloudera continues to tout itself as a platform for machine learning. Everyone who ever knew anything about AI/ML left the company. Marketing just keeps on fucking that chicken.
Feature Labs sought to
Monetize feature engine-
neering but they failed.
Here are three predictions for 2020:
H2O.ai will add multi-user capability to Driverless AI. This will surprise customers who assumed that enterprise software can support more than one user at a time.
Customers who try Amazon Web Services’ SageMaker Studio will discover that it’s JupyterLab repackaged with the AWS brand. Perfectly legal for AWS to do that, but kind of lazy.
Alteryx Assisted Modeling will exit beta in Q1 or Q2. Alteryx will still be good for data prep and nothing more.