Top Machine Learning Companies

Machine learning companies have emerged as key players in enterprise IT over the few years. Why? Enterprise leaders have realized the value of having software that is capable of learning on its own without human intervention. Today, machine learning (ML) capabilities are baked into many different kinds of enterprise software. This type of artificial intelligence (AI) powers everything from recommendation engines to medical diagnostic software to cybersecurity tools to self-driving cars.

As machine learning has become ubiquitous, many organizations have begun building in-house data science teams. Many of these teams focus on analyzing business data to generate valuable insights, while others are incorporating machine learning capabilities into their company’s products or using advanced algorithms to solve industry-specific problems.

For these types of purposes, organizations often go looking for a general-purpose machine learning platform that can solve a wide variety of needs. The ten vendors on this list offer these sorts of ML tools and deserve consideration if you are in the market for ML software.

How to Select a Machine Learning Vendor

The capabilities of machine learning platforms vary widely, and no one ML tool is going to be right for every use case. The selection process should involve carefully taking stock of your organization’s needs — now and in the near future — and finding the best fit. Here are some questions to consider:

• Who will be using the machine learning software? Does your team include experienced data scientists or business analysts with less experience? Or both? What languages and tools do they already know, and which platforms are they likely to learn quickly?
• How do you want to deploy your solution? Some ML tools run on public cloud services, some are delivered as software as a service, and some can be deployed on your own servers. You’ll need to find the option that meets your security and governance needs while providing the lowest total cost of ownership.
• Where does your data reside? You’ll need to make sure that the ML platform you choose will be able to ingest data from your sources. If you already store a lot of your data in a particular public cloud, it might make sense to choose an ML service that runs on the same cloud.
• Do you need data cleansing, preparation and management features? Some platforms offer end-to-end capabilities while others are more narrowly focused on machine learning. Consider what tools you already use and what kind of platform will fit in well with your current workflow.
• Do you need continuous integration (CI), continuous deployment (CD) or MLOps capabilities? Some of the tools listed below will fit better into modern DevOps-style environment than others. Look for a tool with workflows that match the way your team works.
• Are you willing to pay more for faster results? Machine learning platforms with advanced automation capabilities, templates and easy-to-use interfaces often cost more, but it might be worth the expense if your ML project starts returning valuable insights more quickly. You need to consider the optimum balance of cost and productivity for your organization.

With those questions in mind, here are ten of today’s best machine learning vendors, with pros and cons for each:

Best Machine Learning Companies

Jump to:

• • Amazon Web Services
  • Databricks
  • Dataiku
  • Google Cloud
  • IBM
  • MathWorks
  • Microsoft Azure
  • RapidMiner
  • SAS
  • TIBCO

Amazon Web Services

Launched in 2006, Amazon Web Services (AWS) was an early pioneer of cloud computing and continues to be the world’s largest provider of infrastructure as a services (IaaS), platform as a service (PaaS), and private cloud services. A subsidiary of Amazon, the company reported sales of $10.8 billion in the second quarter of 2020 and $21.0 billion through the first half of the year, which analysts says is about a third of the cloud infrastructure market. Many large enterprises use AWS to store at least some of their data, which gives the company an advantage when organizations are looking for a ML provider.

Amazon’s machine learning services center around its flagship SageMaker line of services. They include the SageMaker Ground Truth tool for building managing data sets, SageMaker Studio IDE, SageMaker Autopilot for building and training models, Augmented AI for human review of predictions, and much more. Its customers are some of the most well-respected users of machine learning, such as Intuit, CapitalOne, Siemens, FICO, Kia, Formula1, PWC, Tinder, Yelp, the NFL, Netflix and Pinterest.

AWS offers a free tier that companies just getting started with ML can use for two months. After that, pricing depends on the services used, data center location, instance size (the type of computing resources used) and the number of hours. The prices are all listed on the site, and AWS offers a tool for estimating total costs.

Pros

• Companies that already store data in AWS or use other AWS services (and many enterprises fit this category) may find it convenient to also use AWS for their machine learning needs.
• AWS offers an all-in-one approach that allows data scientists to do everything they need to do all in the same service.
• The upfront pricing lets you know how much everything is going to cost, and you only pay for what you use.

Cons

• You will need to keep tabs on how much you are using AWS services or you may end up with an unexpectedly large bill.
• Some customers complain that the documentation for SageMaker isn’t as complete as it could be.
• The service requires coding expertise and advanced data science knowledge; it doesn’t have the drag-and-drop interface that some other ML vendors offer.

Databricks

Founded in 2013 by several of the people behind the Apache Spark project, Databricks is a pure-play data science and machine learning startup. Its customers include Comcast, Conde Nast, H&M, Regeneron, Nationwide and Showtime. Headquartered in San Francisco, California, it has raised an estimated $897 million in funding.

Databricks’ Unified Data Analytics Platform includes its MLflow-based Data Science Workspace and its Apache Spark-based Unified Data Service, as well as its Redash visualization and dashboarding tool. It runs on AWS or Microsoft Azure, and it integrates with many popular business intelligence tools, including Tableau, Qlik, Power BI, Looker, Mode, TIBCO Spotfire and ThoughtSpot.

The company has different offerings for AWS and Microsoft Azure, and each comes in several different tiers. Pricing is available on the company’s website, and a free 14-day trial is also available. There’s also a free Community Edition with a smaller feature set.

Pros

• If you are considering using Apache Spark in production, Databricks is a great way to get access to all of Spark’s features with the service and support that enterprises need.
• It supports multiple languages, including Python, R and Scala.
• Databricks gets rave reviews for its customer service.

Cons

• Databricks can be expensive, depending on which plan you choose, because you will need to pay for AWS or Azure cloud instances as well as the Databricks service.
• The service is sometimes slow, depending on how much data you have and whether your compute instances are sized properly for your jobs.
• The company’s documentation isn’t the best, and it gets particularly low marks for its search capabilities.

Dataiku

Anther data science and machine learning pure-play, Dataiku was founded in 2013 in Paris, France. In late 2019, the startup announced that it had achieved “unicorn” status with a valuation of $1.4 billion. Its customers include GE, Sephora, Unilever, Ubisoft, Palo Alto Networks, L’Oreal, Capgemini, and Les Schwab Tires. It employs more than 400 people in New York, Paris, London, Munich, Sydney and Singapore.

Dataiku DSS emphasizes collaboration and self-service capabilities. It incorporates both notebooks and a drag-and-drop interface, as well as visual data preparation tools, modeling tools and dashboarding capabilities. And it supports Python, R, Spark, Scala, Hive and more.

Dataiku comes in Discover, Business and Enterprise editions, and the company also offers a hosted online trial and a free edition. Pricing is available on request.

Pros

• Dataiku’s visual interface makes it easier to use than many other machine learning tools.
• The platform is highly scalable.
• The workflows are very flexible and allow you to mix different languages within the same project.

Cons

• Dataiku DSS doesn’t integrate with as many other platforms as some of the other machine learning options.
• You’ll need to contact the company to get a price.
• The free version is not very full-featured.

Google Cloud

Currently the third largest cloud infrastructure provider behind AWS and Microsoft Azure, Google Cloud is Google’s public cloud computing service, including its G Suite cloud-based productivity tools. In the second quarter of 2020, it generated $3 billion in revenue, a 43 percent year-over-year increase. Its customers include Nintendo, PayPal, Macy’s, Spotify, The Home Depot, The New York Times, Toyota, Airbus, FCA, Target and many others.

Google Cloud’s machine learning services include its AI Platform, Cloud AutoML, Deep Learning Containers and TensorFlow Enterprise. All these services draw of Google’s expertise as one of the largest users of machine learning technology and its research into TensorFlow and AutoML. It offers services for every aspect of the ML pipeline, including continuous integration (CI), continuous delivery (CD) and MLOps capabilities.

Pricing for the various Google machine learning services is available on the Google Cloud website and tends to be somewhat lower than other cloud machine learning services. Google also offers per-second pricing, which helps keep costs low.

Pros

• Google has a lot of machine learning expertise, which it integrates into its ML cloud services.
• If you already use other Google Cloud services, its machine learning services could be a very convenient option.
• Google Cloud offers very affordable pricing.

Cons

• Google Cloud does not support hybrid cloud models.
• Some users complain that the AutoML functionality is too opaque, and they don’t understand what is happening under the hood.
• Some of Google Cloud’s ML services are not as flexible as other options on the market.

IBM

Founded in 1911, IBM is one of the oldest and most well-respected companies in the technology industry. Its headquarters is in Armonk, New York, and in its most recent quarterly report, it announced $18.1 billion in revenue. Its total quarterly cloud revenue, which includes IaaS and software as a service (SaaS) was $6.3 billion.

One of the early pioneers of artificial intelligence and machine learning, IBM made early headlines with its Watson AI platform. It continues to sell a host of AI and ML services under the Watson brand name. Its Watson Machine Learning product integrates with other Watson tools and supports hybrid and multi-cloud environments. You can also deploy it on your own servers.

IBM offers a wide range of different pricing options for its Watson services, including a free tier. Pay-as-you-go pricing details are available on the website, but for subscription pricing (which may be lower than pay-as-you-go) or deploying on your own servers, you will need to contact the company.

Pros

• IBM’s Watson Machine Learning is one of the few ML services that you can deploy on IBM’s cloud, another cloud service, your own servers or any combination of the three.
• IBM tries to help companies accelerate their time to value on ML projects, and it boasts that its ML services can boost productivity by 40 percent.
• The company has long been an industry leader in AI and offers an extensive list of related services.

Cons

• Some customers have complained that deploying the solution took longer than expected.
• Like many ML tools, it requires some data science expertise to use the service to its full capabilities.
• Some customers ran into bugs when deploying on their own servers.

MathWorks

Founded in 1984, MathWorks is a privately held company headquartered in Natick, Massachusetts. In 2019, it generated more than $1 billion in revenue, and it has more than 4 million users. Its products have long been a favorite in academia, and its customers include institutions like Harvard, MIT, Stanford, Carnegie Mellon and many other prestigious universities. It also has a long list of enterprise users that includes Boeing, Airbus, NASA, Ford, GM, Tesla, Pfizer, Nokia, apple, Intel, JP Morgan, Amazon, Facebook and Google.

Originally created in the 1970s, MATLAB is one of the oldest products on this list. While most of the other began as analytics tools, MATLAB began as a tool for mathematicians, scientists and engineers. However, the same software that was good at handling advanced mathematics turned out to also be good at machine learning algorithms. It has a Statistics and Machine Learning Toolbox and a Deep Learning Toolbox that can be very useful for data scientists.

MathWorks offers a free 30-day trial of MATLAB. After that, a standard perpetual license costs $2,150 (or $860 per year). Less expensive license types are available for education, home and student users. Also, you can choose to download the software or use the online version.

Pros

• MathWorks’ products are widely used in academia, so recent graduates and data scientists coming from university settings are often familiar with them.
• It’s a good option for writing machine learning code that will be embedded in other software or on devices.
• MATLAB is highly scalable and includes parallel processing capabilities.

Cons

• MATLAB isn’t easy for ordinary businesspeople to use; it was designed for mathematicians, scientists and engineers.
• MathWorks does not offer pay-per-use pricing.
• To use MATLAB, you will need to learn the MATLAB language, as it does not support other popular languages like Python and R.

Microsoft Azure

First released in 2010, Microsoft Azure is currently the second largest cloud infrastructure vendor, with approximately 18 percent of the market. In its most recent quarterly filing, Microsoft reported that its Intelligent Cloud group, which includes Azure, generated $13.4 billion in revenue. It did not break out numbers for Azure but did reveal that Azure sales increased 47 percent year-over-year.

Azure’s Machine Learning service includes both code-based and drag-and-drop interfaces, as well as automation and support for MLOps. It supports a variety of open source tools, including MLflow, Kubeflow, ONNX, PyTorch, TensorFlow, Python and R. It also incorporates tools for detecting bias and managing fairness.

Azure Machine Learning comes in two different flavors: Basic and Enterprise. Pricing varies widely depending on the type of instances you use, but Microsoft does publish the price list and provides a pricing calculator. Microsoft also offers a free account that comes with a $200 credit that could be used toward Machine Learning, and some of its Cognitive Services are free for 12 months.

Pros

• The Microsoft service is designed to meet the needs of both advanced users and beginners.
• The upfront pricing makes it possible to estimate costs accurately, and it is generally very affordable.
• The service is convenient for organizations that use other Microsoft Azure services.

Cons

• Azure Machine Learning lacks some of the integrations to BI tools and other applications that some enterprises need.
• Some customers say that they wish the service had more R-based models.
• As with other ML services from major cloud vendors, you will need to keep an eye on your usage so that you don’t end up with an unexpectedly high bill.

RapidMiner

Founded in 2006, RapidMiner is a privately held data science, artificial intelligence and machine learning vendor. It boasts more than 40,000 customers including Lufthansa, Transport for London, Daimler, Mobilkom, PayPal and others.

The RapidMiner platform includes its Studio, Go, Notebooks, AI Hub and Automated Data Science products. It is both open source and extensible, and it promises full transparency. It also aims to provide very fast results for both beginners and advanced users.

The RapidMiner Studio Free Edition is available for free for limited usage. The Professional version starts at $7,500 per user per year, and the enterprise version starts at $15,000 per user per year. The RapidMiner AI Hub costs $54,000 per year.

Pros

• Very easy to use, RapidMiner is a good option for data science beginners.
• Thanks to its open source development model, RapidMiner offers excellent transparency, so users can understand what is happening with their data.
• The community forum provides excellent help.

Cons

• The tools do not offer as much flexibility as some of the other options available.
• It doesn’t have the pay-per-use pricing that some of the cloud vendors offer.
• People running complex models sometimes report bugs and slow performance.

SAS

Also known as SAS Institute, SAS is one of the world’s largest analytics software vendors. In fact, it is the world’s largest privately held software vendor of any kind. Founded in 1976, it grew out of a statistical analysis program at North Carolina State University and is currently headquartered in Cary, North Carolina. In 2019, it reported $3.1 billion in revenue, and it currently has around 14,000 employees.

Many of the SAS products would be helpful for machine learning, but the most relevant may be its SAS Visual Data Mining and Machine Learning software. Its key features include automated insights and interpretability, automated feature engineering and modeling, a public API for automated modeling, easy-to-use analytics, network analytics, deep learning with Python and ONNX support, integrated data preparation and in-memory processing. It is part of the larger SAS Viya suite.

A 30-day free trial is available. Pricing is available on request.

Pros

• The SAS solutions are some of the most full-featured available on the market.
• SAS makes extensive use of automation, which can simplify the process of using machine learning and speed up the time to value.
• SAS software gets excellent reviews from analysts and users.

Cons

• SAS offers so many different machine learning and analytics products that it can be difficult to select the right one(s) for your needs.
• In some cases, the SAS products are considerably more expensive than competing ones.
• Some developers say that working with SAS takes longer than with other machine learning tools.

TIBCO

Founded in 1997, TIBCO sells a range of software products related to data integration, data management and analytics. Its customers include Caesars Entertainment, Mercedes-AMG Petronas Formula One Team, Bayer, Blendtec, Campari Group, General Mills, JetBlue, Equifax, Fannie Mae, Macy’s, NASA, Panera Bread and United Airlines. It is privately held and headquartered in Palo Alto, California.

TIBCO’s primary machine learning product is TIBCO Data Science. It offers features like data preparation, model building, pre-built templates, version control, auditability, AutoML, embedded Jupyter Notebooks, and more. It also integrates with TIBCO Spotfire, the company’s flagship analytics platform, which also has some ML capabilities.

TIBCO Data Science comes in four different versions: Statistica, Team Studio, for AWS, and the Students and Academics edition. The Statistica and AWS versions offer free trials. Pricing for the Students & Academics version is available online, and pricing for the other versions is available on request.

Pros

• TIBCO’s data integration expertise makes it easier to bring data in from external sources.
• The platform also has excellent data preparation capabilities.
• The tools also get good reviews for their ease of use.

Cons

• The TIBCO tools can be expensive, especially for smaller organizations.
• Some customers say that TIBCO doesn’t update its software as frequently as they would like.
• TIBCO lacks CI/CD or MLOps capabilities.

Machine Learning Vendor Comparison Table