Thought Leaders in Artificial Intelligence: David Talby, CTO of John Snow Labs (Part 2)

Sramana Mitra: Talk about customers that you are currently working with and also customers that you would like to work with. 

David Talby: We are most famous for our work on natural language processing in the stock NLP library. In terms of customers, we work in the health care and life science sector. The last NLP industry survey done in September was done by Gradient Flow. It shows that we have a 54% share of all the healthcare AI teams that use NLP.

Some of the public customers that we have are some of the largest pharmaceutical companies. Roche, Novartis, and Engine are some of our customers. Healthcare systems like Kaiser and Providence health are also our customers.

In the healthcare IT sector, Mckesson, Highmark Health, and GE Healthcare are also our customers. We target data science teams. They can be either in pharmaceutical companies in healthcare systems or startup healthcare IT companies. Most recently, we are looking to work with medical texts. 

Sramana Mitra: You talked about various segments of customers. Could you double-click down and talk about the use cases in these different types of customers? Talk me through the problems that you are solving.

David Talby: It does seem like we sell to different kinds of companies, but usually we do sell to the same kind of team. We sell to teams who are implementing data science in AI within the healthcare setting. The reason why we can sell to software companies, pharma companies, and healthcare companies is that we don’t sell directly to clinicians.

We are selling to data science teams or software engineering teams. In terms of use cases on the pharma side, there are a number of use cases where we are analyzing and understanding texts in context. One common thing that we do is involve data and evidence.

For example, if you are dealing with cancer patients, you deal with the disease for several years. You have tens of thousands of pages of texts that are accumulated. These include all the progress notes, pathology reports, radiology reports, and lab reports, to name a few.

Several use cases become possible if you can automatically read them and build the timeline of the patient. For example, matching patients to clinical trials is a fundamental NLP problem. You need to be able to read the entire history of the patient. You need to be able to read the entire history of the patient. You need to be able to read the inclusion and exclusion criteria of the trials and you need to match them.

Another use case is around clinical decision support. It determines the clinical guidelines for an oncologist when they are looking at the patient. This is a domain-specific thing. We need to know specific things. It’s not just questions like what is the cancer and at what stage.

It’s questions like what is the tumor, what is the histology, is it metastatic, are there other comorbidities, should we operate or not, and should we do chemo or not. We need to know the exact details of what is going on. The other thing is just building a lot of databases of patient histories so that we can produce medical evidence.

Right now, COVID is the thing that everyone is talking about. It is critical to know as early as possible how effective the vaccines are. What side effects are there? The news today is all about COVID, but if you think about it, this is critical to think about when you are dealing with cancer, autism, Alzheimer’s, and other diseases.

There is the whole concept of real-world evidence and for the first time in history, we have all this data digitally. We digitized this data and it’s all unscrutinizeable free text in super domain-specific jargons. The question is how we can automatically extract the data at scale.

This segment is part 2 in the series : Thought Leaders in Artificial Intelligence: David Talby, CTO of John Snow Labs
1 2 3 4 5 6 7