How IBM’s data science team is accelerating user AI projects

Many computer stores have machine learning and AI projects underway, or are planning to launch one. But many struggle to find their first use case or properly mobilize technical and human resources to ensure ROI. It’s Seth Dobrin’s job, as leader of IBM’s newly formed elite data science team, to help enterprise users achieve these goals.

Dobrin, officially vice president and chief data officer of IBM Analytics, discussed his team’s experiences and insights with their early engagements that included more than 30 clients.

What surprises you the most when working with users on their first AI or machine learning projects?

Seth Dobrin: One thing is that most companies will hire a bunch of data scientists and tell them to go do a bunch of “data science” stuff. They won’t have a strategy or value proposition, and they won’t know how to operationalize data science in the context of the business.

[By comparison,] a manufacturing company wouldn’t buy a new piece of equipment that didn’t have a return on investment proposition. My fear is that in two years these companies will be saying, “We invested in this data science, and we haven’t generated any value from it.” Part of my job is to help them create a value proposition and develop a strategy around it.

How do you start this process?

Dobrin: We walk them through our methodology for identifying a specific use case, then break it down into component models. Most use cases worth realizing are not a single pattern, but include several. Then, together, we build everything they need for these models in an agile way. At the end of the engagement, which does not exceed 90 days, we leave them a working model. But, more importantly, we taught them to [data science-based projects] within the framework of the company.

Once you’ve agreed on a use case, what tools does IBM bring to the table?

Dobrin: For up to 90 days, customers have free access to our products and services. We have a data science platform [Data Science Experience Local] which can be deployed on any cloud, including that of Amazon or Google. We have Watson Studio on IBM Cloud. Depending on their use case, we teach them how to use these tools.

With more advanced customers with shorter engagements, we show them the added value of our tools, like the ability to deploy an API, as well as automatically retrain their models without having to consume their data scientists’ time with it. We also show them all the collaboration features, so they can automatically connect to data, as well as perform data exploration without having to write any code.

What tools do users need to have to work with you on AI projects, especially if they’re not True Blue shops with things like IBM Cloud?

Dobrin: They have to be trained to learn how to access the platforms, which they can access for free for 90 days. But they still have to provide the hardware on which we install the software, and we will defend the environment. Some users see [our platform] adds much more value over their own or off-the-shelf components they might purchase.

So are you sticking with them until they launch their first AI or machine learning based offering?

We believe the future of data science is in the vein of Python.

Dobrin: Yes, and we’re confident we can do it in 90 days. We have a good plan to make it happen as long as they hire a group to work with us. One of our requirements of our clients is that they should at least match our resources. In the field, I usually assign three or four people to an engagement. Customers initially are adding three or four people, but so far we’re quickly three or four to one, which we take as a sign that we’re adding value.

Typically, what kind of use cases are customers most interested in?

Dobrin: We have engaged with oil companies exploring for oil, retailers looking to optimize the efforts of sales staff, and manufacturers looking to streamline their manufacturing processes. Typically, this helps them build a machine learning or decision optimization solution, so they can build a predictive or prescriptive model that they can [add to] their processes. We don’t do the physical implementation of their software – we just provide an API. If they want someone else to do the implementation, we refer them to IBM’s service organization.

Do you require user organizations to have some proficiency in AI and machine learning, and do you generally find there are enough of such people?

Dobrin: We prefer that [users’ data science teams] have some ability with a programming language. We believe the future of data science is in the vein of Python, so we really like people who know Python. Data science is not something you can just [quickly] teach people. It takes more than three months to teach people to do data science.