I don’t quite remember when I have heard about dataiku for the 1st time but when I started my consulting activity, focused on AI, it was clear to me I had to make a good selection and chose an AI platform so I could bring efficiency to my clients and consulting projects.
I first took into consideration the fact that a lot of the underutilized data in companies is tabular data. I could have chosen scikit-learn then, but being a 1 person company I did not want to invest time in mastering each and every components involved in serving the machine learning value (plus admittedly I’m no longer a heavy daily coder).
I wanted to get quickly from project start to a predictive POC (proof of concept) in matters of days (not weeks) so medium businesses could leverage their data’s value potential quickly. Dataiku embraces all the data wrangling involved in EDA (exploratory data analysis) and it is nicely centered around building and evaluating machine learning models.
The platform also natively supports the iterative nature of the AI projects (also referred to as MLOps) and brings consistency all the way from start to deployment.
Dataiku’s architecture is based on proven open source components serving big data and machine learning products worldwide, so I know it will scale too (and not require complex debug or complete redesign in the midst of program’s execution).
Then, teamwork is another key attribute that was very influential in choosing dataiku. I have used flow graphical representations in many of my previous work experience. A flow view is a very efficient way to communicate the various processing steps involved in any data project, yet support abstracting its complexity by using composition.
It is also no secret low code or no code platforms are trendy and there is a good reason for that: specialists are scarce resources, whether developers or data scientists. We have lived in the Zettabyte Era for years already, and there are basically more use cases and data, than companies have specialists to handle it all by themselves.
So having the right tool to federate and make productive the data citizens is key (a data citizen is just everybody in your company, the non-specialists). By sharing the same platform, the numerous data citizens will efficiently collaborate and help the data scientist make sense of the various workflows the company has developed and responsible for generating all that data. This close collaboration around data and AI output facilitates business value creation and largely avoid latent inefficiencies caused by tools disparities or access.
Again, being a 1 person company advising companies on many different projects make each project unique. The AI platform I was going to use needed to be complete so connecting to the various enterprise data sources would not be requiring much if any custom work, yet open so it could evolve for specific needs if required.
Dataiku is also complete in the various models you can choose from, making it super straightforward to leverage AutoML on all these options, yet open for extensibility to bring in any necessary project specific AI code.
The set of data visualizations included in the platform fits very well with the various actors need to interact and discuss most the various challenges in an AI project (prediction models’ performance, explainability, data drift).
Also, serving multiple clients, I could not picture myself being a bottleneck to my customers. Dataiku comes with free online training courses supporting my customers’ employees learning with my recommendation and/or at their own pace.
There may be other reasons for selecting and using Dataiku (governance, security, …), I have long wanted to spend time sharing why I took on becoming a Dataiku partner, to better serve my customers needs.
I hope you enjoyed this article and motivated you to download and try out Dataiku ASAP.
Please don't hesitate to reach out to Hervé @ biZNov if you have questions or just to let me know if there other subjects you’d like to be treated on this blog.
And don't forget to spread the information if you enjoyed this blog post, just click on the social network buttons below. Sharing is caring :-)
Comments