This is episode one in the Accern Podcast Series. Here, Accern Co-Founder and CTO Anshul Pandey leads the conversation. He sits down with NVIDIA Customer/Partner Developer Relationship Manager Prabhu Ramamoorthy. Much of the conversation revolves around the idea that unstructured data and NLP (natural language processing) are the epicenter of AI (artificial intelligence). They touch on the future of AI and deep learning. And they look at AI applications in fintech. Finally, they consider what makes a great product in AI, ML, and NLP.
On the future of AI and deep learning
When it comes to the future of AI, we are perhaps only constrained by the creativity of our imaginations. Below, Prabhu lays out the current state of AI and what he is currently excited about.
“The future is very bright. AI for me and my customers means many things. It could mean quant finance, which we call SPC. And there are areas such as extract transform load (ETL) machine learning, which is using non-neural nets. So that's the second area. And the third subset is deep learning AI, what we call neural nets. And deep learning can be used for both structured data as well as unstructured data.
“So we see a world in the future where customers are looking for specialized solutions. Today we speak with digital assistants in English. But tomorrow, it could be in various languages. It has to be customized to different dialects and accents. And we see this vision of a deep learning future where you will have a lot of data, and you will be able to train your models to recognize natural language processing. And we would have a full blown solution for unstructured data as well as real time analytics. This area will continue to become intensive within both hardware and software. And we want to be able to support customers in developing those use cases.
“We are going to lead into a world where every language is going to be supported. Things that were not possible, those applications can be done. And you have to have the user driving it to the end customers.”
On AI applications in FinTech
Artificial intelligence, unstructured data, and NLP are changing banking. In particular, FinTechs are deeply involved in AI, in part because it is a competitive advantage. Prabhu explains below.
“In the past, there was a limited application, you had something called a core banking system. That stored banking transactions, and solutions would be built on top of that. But what we are seeing now is that customers are going beyond. For a banking or an asset manager to be relevant,they are competing against the fintechs. And fintechs are serving a bigger portion. Those who are underbanked. So core banks, for example, Goldman Sachs or Bank of America, are now competing with these biggest customers, and they have to ensure that the latest generation, millennials, adopt it.
“The trend that we see happening is everybody is looking at large scale greenfield projects in AI. And when they mean AI greenfield projects, these are projects related to unstructured data. It could be a chatbot. Or it could be a natural language processing solution which answers the customers.
“This is giving a competitive advantage to that interface customer, so that they can capture more customers. There are two angles to it. Organizations that adopt this gain a competitive advantage. And we also see that organizations that do not adopt this are at a severe disadvantage, and there is an existential threat to their business. Where they probably risk losing out to a competent and successful fintech.”
On great products for unstructured data and NLP solutions
With so much need for products that can solve problems of unstructured data and NLP, what makes a great product in this space?
Prabhu says, “I am always on the business side. You have to take a holistic look at every project, and it has to be connected to a use case. You can do research for the research sake of it. And as you probably know, there are a lot of AI projects in that category, where it's about enhancing the ecosystem. So that research has to be done.
“But when we come to the end customers, like enterprise financial services customers, the one thing that I see is in order to build a winning product, you have to be able to tie it to a holistic use case. You have to look at the big picture. Understand that that product is helping out in some critical use cases. The cost of AI has become a lot in terms of compute, hardware, and software. You have to hire data scientists. So it obviously helps when you have a product in hand that is able to generate return on investment for the end customer.
“I'll take an example like document understanding. And it's providing an ROI. It's reading mortgage contracts. Key is to put yourself in the mind of the business. The product has to be able to help in an ROI. And that discounts the cost of AI. And now you go there, and you go and start solving business problems, the AI project starts paying for itself. Then you could use an Agile process to increase and innovate the product.
“I believe that's like the way we have been seeing companies grow. Organizations typically start with an AI Center of Excellence, but then they quickly try to apply it working with business to apply to specific use cases. There is a place for both build-your-own as well as buy-with-partners. So there are two ways where an AI process could be done. In both areas, the customer has to test it out in actual use cases. I would like them to prioritize the ROI and apply it and once they have been in one area. Go to the next area, tackle the next area, and increase the product innovatively.”