Agent Workforce Platform for AWS

My main use case for DataRobot is building machine learning models.

For a specific example of a project I've built with DataRobot, we were trying to build a causal inference model to understand customer churn, and I used DataRobot because it was easy to understand. Once you have data, you put it in, and it builds a variety of ensemble models and machine learning models. It provides a very easy interactive dashboard.

DataRobot has positively impacted my organization by reducing the model building time from around one week to two days.

Besides the time savings, I have noticed improved accuracy as one of the positive outcomes. Whenever we did the confusion matrix, the sensitivity and specificity improved significantly because sometimes we miss some variables and techniques in feature selection. DataRobot's inbuilt capability identifies which features are useful, leading to a 5% improvement in accuracy.

I chose DataRobot for this task over other platforms because of its usability and features. Once you have any kind of data, you put it in, and it has each step outlined. It tells you what to do next, what is required, and provides the results in a very good manner.

For data cleaning, DataRobot is really good. If you make errors in data cleaning while building a machine learning model, it helps you and provides a wide variety of machine learning models that you may not have considered. It gives better results, and it is very easy to export the data and results to show to leadership.

The best features DataRobot offers are the results that allow you to identify how a model is built and show the interpretability of different models.

Regarding interpretability, DataRobot helps my team understand the models by providing explanations of why a certain model is considered, how the data is involved, why the model is chosen, and what the use case is. It defines everything, letting you know why it is choosing that.

Regarding DataRobot's AI capabilities, its governance and security are good. I have used DataRobot for financial services, where data undergoes a lot of scrutiny. DataRobot takes care of all the data and does not share personally identifiable information with anyone.

DataRobot's accuracy and reliability of output are solid. Similar to any tool, I should be cognizant of limitations, but with DataRobot, since you are interacting and playing around with it, I believe that 95% of the time, we can trust the results it provides because we know what is happening, thus ensuring accuracy.

One way DataRobot can be improved is that for some non-technical people, it is very hard to understand some terminologies. Providing more training would be beneficial because it requires substantial training. Without training, it is difficult to use.

In terms of needed improvements, documentation and support are crucial. We have support, but it can be slow. The documentation should clearly outline each step and what to do next. If something becomes problematic, there should be a help button immediately available to provide answers, rather than waiting on the support team.

I have been using DataRobot for a year.

In my experience, DataRobot is stable.

DataRobot's scalability is good. It can handle multiple tasks, not only machine building models but also excels in text analytics.

The customer support rating I would give is a six because sometimes it is hard to get immediate service.

Before using DataRobot, we were using Python for end-to-end processes, and in a previous organization, we used Enterprise Miner to build models. We switched because we wanted to do something scalable.

I have seen a return on investment with DataRobot in terms of time saved and fewer employees required. Previously, when building models and working on multiple projects, one person could not handle it alone, necessitating hiring another person. Now, one person can manage it because DataRobot takes care of everything behind the scenes, improving the model building time from one week to two days.

Before choosing DataRobot, we evaluated options including Enterprise Miner and other small solutions, though I don't remember the names currently. DataRobot stood out as one of the tools we decided to use.

On a scale of 1 to 10, I rate DataRobot a seven.

I chose 7 out of 10 because of some usability difficulties for non-technical members. I want to see DataRobot become usable by anyone, whether technical or non-technical, to interpret the results.

My advice for others looking into using DataRobot is to use it with your best knowledge because it is very powerful. Whenever you are using it, try to understand what the output has provided. Start with the documentation to grasp its capabilities, as it offers a variety of features that need to be understood clearly. Documentation is important once you start using it, and attending the training that DataRobot provides will be very helpful.