AutoML aims to democratize the integration of artificial intelligence within business applications by making it more accessible to users with varying levels of expertise. It reduces the time and resources required to build and deploy machine learning models, which in turn has several benefits for the industry. In this post we evaluate emerging AutoML platforms covering a few relevant industry use cases.
What is an AutoML Platform?
An AutoML platform offers a suite of streamlined and prepackaged machine-learning models to accelerate the end-to-end machine-learning workflow for a specific real-world problem.
AutoML platforms typically rely on foundational machine learning models as the building blocks for their automation processes. These foundational models serve as the basis for the automated model selection, hyper parameter optimization, and evaluation of steps in the AutoML workflow. Depending on the type of use case, different models, such as classification models, prediction models, feature discovery models, or generative models are offered to address specific problems.
Different Types of AutoML Platforms
The main differentiator for any AutoML platform is it’s user experience which wraps around the complexity of the underlying models and presents a simple and intuitive interface for users. Based on the primary user interface offered to the end users, AutoML platforms can be broadly categorized under the following types:
- API Based: For very narrow, task-specific use cases of ML, an API call is a good option. For example, use cases such as a general image or text classification can leverage a hosted ML model, triggered through an API. The AutoML platforms wrap these models around an API interface and users follow the standard API invocation mechanism to pass the input data and get the results returned by the hosted model.
- Low-code, No-Code Interface Based: For more complex scenarios, a low-code or no-code visual interface offers a lot of options to design a custom workflow that can be tweaked at various stages for feature extraction, hyperparameter tuning, and model evaluation. The visual interface is also augmented with data visualization to analyze the model outcome and for further data analysis.
- RPA Based: A Robotic Process Automation based AutoML platform provides guided workflows for the user to walk through the machine learning process step by step. These workflows are specifically devised for an industry use case. Such platforms may also have a no-code, low-code interface to allow the users to visually monitor and tweak the workflow.
- Component Based: These AutoML platforms offer a plugin or a widget that can be integrated with any website or application, and controlled from the backend to offer certain ML functionality. The most common example is a chat bot widget that interacts with website visitors but its output is generated based on the per-trained data fed from the platform’s admin backend.
Benefits of Using AutoML Platforms
- Increased productivity and decreased time to market: AutoML automates many manual and repetitive tasks involved in the machine learning process, such as data preprocessing, feature engineering, and comparison of models for optimal accuracy. This results in faster deployment and requires less technical expertise. Due to a streamlined and iterative process outlined by the AutoML platform, it becomes easier to perform rapid experiments, validations, and refinement, leading to a faster real-world deployment.
- Reduced dependency on Data Scientists: AutoML platforms abstract away the complexities of machine learning algorithms and techniques, making them accessible to users with less technical expertise, compared to highly skilled data scientists. It allows domain experts, business analysts, and citizen data scientists to leverage machine learning effectively without extensive coding or data science knowledge.
- Democratization of machine learning capabilities: With AutoML, business users can take a more hands-on approach to machine learning, experimenting with different models, algorithms, and parameters to address business challenges and opportunities. This empowerment democratizes decision-making and innovation by enabling business users to drive AI initiatives and generate data-driven insights independently. Additionally, AutoML platforms take care of model deployment, maintenance, and operational aspects to handle the scale and manage the performance of model outcomes, thereby alleviating the end users from the day to day maintenance hassles.
Top 10 Emerging AutoML Platforms to Watch Out
1. Obviously.ai
Type: Low-code, no-code based
Obviously.ai provides a variety of pre-built machine learning algorithms, covering a wide range of machine learning techniques including classification, regression, and clustering. It is suited for business use cases around prediction based modeling on tabular or time-series data. Example use cases include revenue forecasting, sales prediction, and credit risk scoring. Check out this interesting use case of employee attrition prediction using Obviously.ai.
2. Levity
Type: RPA based
Levity is a workflow automation platform that incorporates machine learning capabilities within business process workflows with support for connecting to hundreds of apps through Zapier, Make, or APIs for seamless integration. Levity’s key features include a best-in-class workflow integration with existing business apps, an AI flow builder for custom workflows, and continuous re-training through human-in-the-loop. Here is a great example of how to use this platform for product image classification.
3. Nyckel
Type: API based
Nyckel is a no-frills AutoML platform that offers a simple API interface for performing most mundane classification tasks on images and text. With a straightforward UI and a bunch of APIs, anyone can use this platform to build a machine-learning classification function within 5 minutes to upload, label, train, and deploy a model. Have a look at this use case on comment moderation to experience the ease of using this platform for text classification.
4. Simplifai.ai
Type: RPA based
Simplifai.ai offers automated machine learning pipelines to build end-to-end email, document, and chat workflows within an enterprise. It offers custom-built GPT models for industry-specific use cases like Insurance. It also guarantees best-in-class compliance to safeguard customer data used in the platform. Here is a great insight on delivering customer service using AI workflows powered by Simplifai.ai.
5. dotData
Type: Low code, no code, and RPA based
dotData accelerates the development of AI-powered business workflows through automated feature engineering capability to expedite model deployment. The platform synergizes AI with BI (Business Intelligence) and takes care of data wrangling, manipulations, and machine learning automation, including feature cleansing, to deliver state-of-the-art machine learning models for both mid-sized and large enterprises. Take a look at this point of view on customer churn prediction in insurance.
6. Modulos
Type: Low code, no code based
Modulos platform offers full lifecycle, machine learning implementation for enterprises to execute and review the performance of AI initiatives. The Data CoPilot® suite from Modulos offers an AI digital assistant-like interface to perform data analytics through exploratory means and visualization. To get a deeper understanding of this platform’s capabilities, have a look at this use case for fast and efficient image classification.
7. Appian
Type: Low code, no code and RPA based
Appian is an advanced AutoML platform to rapidly develop custom applications with a visual interface and pre-built development modules to reduce the time required for application development. The platform’s core features include an advanced engine for modeling, modifying, and managing complex processes and business rules. It offers custom AI modeling and a built-in data fabric layer to connect enterprise-wide data. The platform boasts itself of supporting an impressive set of use cases cutting across many industries and some emerging use cases like ESG management.
8. Squark
Type: Low code, no code based
Squark is a low code AI as a Service platform to build and operate end-to-end machine learning workflow that helps enterprises make better decisions to deliver results for their everyday business tasks. It supports connectors with well-known third-party systems such as Salesforce and Snowflake, in addition to ingesting tabular and textual data from Google Sheets. Squark also supports forecasting on time series data.
9. CustomGPT
Type: Component based
CustomGPT is an AutoML (Automated Machine Learning) platform that leverages the GPT-3 and GPT-4 language models to automate various tasks. CustomGPT is built on top of these models and can be used for tasks such as text summarizations, programming code, generative content, and more. It can also be customized to provide detailed explanations tailored to specific use cases, like generating marketing content.
For a comprehensive coverage of AI use cases across different domains and industries, have a look at our AI use case resource page.
