Dataiku for Enterprise AI (Non U.S. Markets)

My main use cases in Dataiku  include ensuring a strong data pipeline ingestion. We have people from data management, so we need to take care of the pipeline, their data quality, data drifting, all these things. We are taking care of it with the Dataiku  rule-based alert systems we have created.

The best feature in Dataiku is that once the data is connected in the underneath layer, it flows exceptionally smoothly if you know how to tweak it. If you don't know, then it will create a mess. If you know how to tweak it and make the data according to your requirement, then it will be good. If you don't know and are trying to learn on the production, then it is a disaster.

I have used Dataiku's AutoML tools. The AutoML tools have helped me on the fly, as you can apply the machine learning models. They are continuously reading your data and then creating the feature enablement. The moment feature enablement has happened, then you can do the model registry on the fly. Those model registries can trigger your new data. Imagine whatever the data test and train that is passed. Your operational data which is coming new every day, then that feature is enabled and it will give the reasonable amount of prediction and reasonable amount of value on the column so that you can utilize those. You can consume those in the application layer.

Dataiku's data source integration flexibility is completely up to the requirement. We are not using it for ourselves. We are using it for business teams, and they are sending the requirement and we are ingesting according to their requirement. The important thing is, imagine raw data is coming A, but they need A plus B plus C multiply by D. All those kinds of enablement we are doing with the help of Dataiku.

Our source system, the core system, is continuously throwing the raw data on the landing layer. Then from the landing layer, we are converting those raw data and making it as a consumption layer, consumable data. With the help of this, we are doing it.

In terms of enhancing collaboration within my team, I would not say Dataiku is the best one because it's so expensive. We are not able to provide it to everyone. There are very few people who have the developer license and are using it. Once the data pipeline is created, then we are directly handing over that data pipeline to our user on the ingestion layer. It is not a very cost-effective solution, I must say, though it is good for developing purposes only.

Pricing can be improved.

I have been using this product for four years.

In my opinion, Dataiku is stable because we know how to use it. There are many unstable things happening, so it's not that only the application is stable or unstable. Even so many other things, we are facing challenges. I cannot only blame one thing.

In terms of stabilization, if my data has no outlier creation in the raw data, then it is quite stable. I would rate it a seven.

For support, I haven't created any support tickets, so I really don't know about it, but it is quite good.

Positive

The initial setup started with HANA . Then they introduced Databricks . When Databricks  got live, then they started giving this license for Dataiku. We got the Dataiku license and learning. Everything went smoothly. Now Databricks is replaced by Snowflake . Even on Snowflake , we can do many things.

It is hard to say if I've seen a return on investment in Dataiku because we are far away from the monetization of the data. There are other teams who are taking care of the monetization. We are not from resource management, so it becomes very hard for us to calculate the ROIC on this at each and every application level. We are not using only Dataiku, we are using many other products.

In my opinion, it is good, not bad. I must say because I'm using many other tools as for a data operating model. It is much better than other tools because it has a clickable solution. Most of our data citizens who really don't know the coding thing can easily do things with the help of the mouse. Most of the things are working fine, so there is nothing to complain about.

Overall, Dataiku is really good. I would rate it an 8 out of 10.

Hybrid Cloud

Amazon Web Services  (AWS )

My primary use case for Dataiku  is for data science, Gen  AI, and data science applications. Our AGN team also uses it for various purposes.

Dataiku  is highly regarded as it is a leader in the Gartner ranking. It offers most of the capabilities required for data science, MLOps, and LLMOps. Integration with public cloud and multiple other platforms is excellent. The product is easy to install and can be maintained by a single expert. It supports good functionalities that are essential in data visualization and responsible AI.

Dataiku's pricing is very high, and commercial transparency is a challenge. Support is also an area needing improvement. More features like LLM security, holographic encryption, and enhanced GPU integration would be beneficial.

I have been familiar with Dataiku for the past four to five years.

I have not encountered any deployment issues. It is very easy to install.

I have not used Dataiku at the level that would allow me to comment on performance latency for a Big Bang environment. However, the product is good, and the output meets our expectations.

Dataiku is fully scalable, and I have not identified any limitations regarding scalability so far.

The technical support from Dataiku is not good. The support team does not provide adequate assistance, and there are concerns about billing requests.

Negative

There are many products available in the market like Converge.io, Domino Data Lab, and ClearML. Dataiku's pricing is not competitive with these solutions.

The initial setup of Dataiku is very easy. A single person, if experienced, can handle the installation and maintenance.

Without a reduction in price, I doubt users will see a return on investment. The market is competitive, and Dataiku must adopt a consumption-based model instead of the current monthly model.

The pricing for Dataiku is very high, which is its biggest downside. The model they follow is not consumption-based, making it expensive.

There are many products in the market like Converge.io, Domino Data Lab, and ClearML.

Overall, Dataiku is a very good product except for the commercial aspect and the support. More features like LLM security and holographic encryption would be appreciated. I would rate the technical support three out of ten due to its current inefficacy. For pricing, on a scale of one to ten, where ten is expensive, I rate it around eight to nine. I rate the overall solution a ten.

Public Cloud

Other

My company sells licenses for both Dataiku and Alteryx, and we have clients who use them. I engage with several companies in telecommunications, retail, and energy to assess how our clients are utilizing these platforms.

The most valuable feature of Dataiku, in my opinion, is the possibility to use Spark, which is a distributed compute engine. This is a feature that is usually appreciated by our customers. 

Additionally, the automation features have been impactful, particularly in the deployment phase, as we use what Dataiku calls deployer nodes. Dataiku primarily enhances the speed at which our customers can develop or train their machine learning models since it is a drag-and-drop platform. Our clients can easily drag and drop components and use them on the spot.

There is room for improvement in terms of allowing for more code-based features. I would love for Dataiku to allow more flexibility with code-based components and provide the possibility to extend it by developing and integrating custom components easily with existing ones.

I have been working with Dataiku for about three years.

I find the pricing of Dataiku quite affordable for our customers, as they are usually large companies. However, it is a pricey solution and I primarily recommend it to bigger companies.

I researched products like Dataiku, Cloudera, and Databricks.

I would give Dataiku an eight out of ten. Although I generally recommend Dataiku, it is mainly suited for companies that can afford it as it is a pricey solution.

Other

I use that IQ since I am preparing cohorts for health investment research.

Traceability and collaboration are essential for me. I have eight or nine engineers working together. Integration with machine learning is also crucial for us. 

Additionally, traceability is vital since I manage many cohorts, and collaboration is key as I have multiple engineers substituting for one another.

I need more experience in the sector, which is health. The license is very expensive. It would be great to have an intermediate license for basic treatments that do not require extensive experience.

I have used the solution for six or seven years.

The solution is scalable. I rate it nine out of ten.

The customer service team is helpful and responsive, more or less on time. I rated them seven out of ten.

Neutral

Deployment should take four or five hours, yet customization takes more time.

Two or three engineers took part in the installation process.

I do not care about financial benefits, however, I am sure they exist. It has supported our compliance with industry regulations one hundred percent.

There are no extra expenses beyond the existing licensing cost.

I work with other tools but mainly with Dataiku , and I also use Python and Azure  Synapse .

The user interface is useful for collaborative tools that allow multiple professionals to work together. 

I rate the overall product as eight out of ten.

Private Cloud

Microsoft Azure 

Dataiku  is an AI platform that we use for oil and gas exploration. Even though I can't provide specific details, this is the primary use case for us.

One of the valuable features of Dataiku  is the workflow capability. It allows us to organize a workflow efficiently. The platform has a visual interface, making it much easier for educated professionals to organize their work. This feature is useful because it simplifies tasks and eliminates the need for a data scientist. If you are knowledgeable about AI, you can directly write using primitive tools like Pantera flow, PyTorch , and Scikit-learn. However, Dataiku makes this process much easier.

One area for improvement is the need for more capabilities similar to those provided by NVIDIA for parallel machine learning training. We still encounter some integration issues.

We have been using Dataiku for three years.

Customer service is somewhat different because Dataiku partners with local industry experts who understand the business better and provide support. It can be challenging to determine the provider of better support, however, overall, the support is good.

Positive

We are only using Dataiku.

I believe the return on investment looks positive.

I considered another option that excels in parallel processing. However, it falls short in other areas. No product is perfect. If these two solutions worked together, it would be advantageous. Unfortunately, one has strengths in certain areas while the other excels in another.

Why not? BHP sold the energy part to a company called Woodside. It has changed because they are now part of Woodside. 

Overall, I rate the product eight out of ten.