The popularity of AutoML solutions has increased in recent years. There are many companies offering a wide range of solutions. These solutions focus on common business problems and tasks frequently pursued by data scientists.
Below is the list of top AutoML solutions. This list is based on AIMultiple, a technology industry analyst.
DataRobot, Dataiku, H2O, Compellon, Enhencer, Akkio, TPOT, dotData, BigML, Prevision.io, TIMi Suite, B2Metric, MLJAR, DMWay, Auto-sklearn, Aible, Auto-WEKA, Tazi.ai, PurePredictive, Caret, Xpanse AI, OptiScorer, Auger.ai
The technology giants are also leveraging their existing infrastructure to push AutoML solutions. They have products to build, deploy and scale ML solutions. Most cloud platforms have dedicated products focused on image recognition and text analysis. They also offer products for building high-performance ML models on structured data. These solutions are quite good in features selection, model selection, and model tuning.
As per ResearchAndMarkets, the AutoML market generated a revenue of $269.6 million in 2019. It is expected to hit $1.5 billion by 2030. It also suggests that cloud-based AutoML solutions are being preferred. As they offer scalability and flexibility to customize solutions.
Benefit of AutoML
Many tasks performed by data scientists are repetitive and time-consuming. This limits the ability of the data science team to work on more business problems. Most data science teams end up focusing only on business-critical issues. Here is how AutoML would empower the data science team
Time-Saving
- AutoML will enable automation of repetitive and manual tasks that are susceptible to human errors
- AutoML will reduce the effort required on data cleaning, exploration, and feature engineering from weeks to days
- AutoML will make the model selection and performance monitoring easy
- Hyper-parameter tuning can be completely automated using AutoML
Level Playing Field
- High performing pre-trained ML model will be accessible to everyone
- AutoML will make machine learning attainable to many small and medium business
- Will make it easy to use unstructured data in making business decisions
- Will increase participation of non-technical users in solving data-oriented problems
Impact of AutoML on Data Science Projects

Automl will create a ripple effect across the life-cycle of a data science project. It will be changing the landscape of data science jobs. Below are typical stages in a data science project. Let’s see the impact of AutoML across these stages.
Business Understanding
The first step in any Data Science project is to well understand the problem. AutoML in general has the least impact in this area. The best AutoML can do is automate some popular and standard data science projects. One example is fraud prediction in banking. The fraud patterns might not change much from one customer to another. It will be easy to template a solution and take it to the market. These pre-built solutions will have the business knowledge incorporated into them.
Will AutoML automate all the business problems? The answer is definitely No! There are many scenarios that can be automated or even replicated. But, an in-depth understanding of a business problem is mandatory.
Data Collection and Cleaning
AutoML products will definitely have access to the data present in the platform. It will be much easier to incorporate new data with AutoML products. For example, In GCP, it is easy to import data into AutoML tables as flat files or using BigQueries. The use of external datasets often produces better results in many problems. To identify the relevant external data, the subject knowledge of a data scientist is a must.
After importing the required dataset, the next step is to clean the data. This step is generally tedious and needs a lot of attention from data scientists. The datasets are generally not clean enough for ML models consumption. AutoML solutions will come in very handy here. We would be able to clean the data a lot quicker.
AutoML will be able to speed up the data collection and cleaning. It will be possible to bring the effort required from weeks to days. But, a data scientist’s domain knowledge will be the key to reach the best solution.
Feature Engineering
This is an iterative process and easily the most time-consuming stage in a data science project. With AutoML it will be easy to implement some feature engineering tasks. Tasks like, normalization, one-hot encoding, binning, formatting can be done with a click of a button.
Can feature engineering be completely automated? The short answer is No. Many tasks performed in feature engineering will become accessible. But, only the data science team can incorporate business insights into the features. Also, for best results, it is often required to look for the unknowns. Making use of the known features and transformations will help only to an extent. To achieve high-performance someone need to look into the unknowns.
Model Building / Insights
The data science projects belong to 2 broad categories. One that involves model building and the other focused on insights or recommendations. The AutoML can be very helpful in projects involving model building. AutoML makes the model selection, tuning, and tracking easy to implement. Thus creating time for the data science team to work on more problems.
The projects that involve insights extraction need business knowledge. With AutoML a lot of focus would shift towards this area. Enabling the data science teams to take up more insight-driven tasks.
Deployment
In many data science projects, deployment has never been easy. There would be challenges moving the models from one environment to other. Any minor difference between the environments like software versions could cause issues. Also, the production environment is usually restricted. Making it difficult to make changes or track the performance of the ML models.
AutoML has shaken up this space. It is now possible to deploy models in a matter of minutes.
What does it mean to Data Scientists?

In the next few years, AutoML solutions will be widely available. Companies of all sizes will have access to the same cutting-edge solutions. To outperform companies would have to focus on problems that can’t be solved by AutoML. This will open up a whole lot of new opportunities for data scientists.
There will be changes to the roles and responsibilities of the data science team. The focus will shift towards more complex issues requiring human expertise and domain knowledge.
Here are some problems that require business knowledge and information about ground reality,
- Insights-driven problems like identifying reasons for customer churn need domain knowledge. There could be a lot of internal and external factors causing the customers to churn. Only better knowledge about the data landscape and domain can produce better results.
- Measuring and tracking the performance of various features/products offered by a company. Business knowledge would be required to define and build metrics.
- Models deployed require constant monitoring. Over time things change and it will impact the model performance. It will require data scientists to fix them.
- Problems that are not clearly defined by the business stakeholders. It requires groundwork to first clearly understand the problem.
AutoML would most likely open up a whole lot of new opportunities to be solved by data scientists. AutoML should not be seen as a tool to make data scientists obsolete. Yes, AutoML will do a lot of tasks currently being done by a data scientist. It could do those tasks much better as well. But, that will lead way to better challenges to be solved by the data science team.
Final Words
The focus for data scientists will shift more to better understanding the problem. AutoML will increase the productivity of the data science team and will empower and not forbid.
To stay connected
- If you like this article and are interested in similar ones, follow me on Medium
- I teach and talk about various data science topics on my YouTube Channel. Subscribe to my channel here.
- Sign up to my email list here for more data science tips and to stay connected with my work