Executive Interview: Atul Arya, Blackstraw.ai


Blackstraw.ai Seeks to Fast-Track
AI Adoption with Smart Building Blocks

AI Trends recently caught up with Atul Arya, CEO and Founder of Blackstraw. The veteran machine learning leader most recently at Nielsen is transforming the AI space using his real-world experience of implementing AI with speed in enterprises. Story by Jeff Orr of AI World.

Q: Hi Atul, can you tell us more about Blackstraw and what it stands for?

A: The tagline we have at our company is “Simplify AI”. Our vision is to fast track AI adoption in enterprises having use cases ripe for prediction or where there is a definitive need of applying learning from old data to new data. We help evaluate the state of data in enterprises and then identify and implement strategic goals of organizations that require AI. The key is to use AI only if AI is needed and when an enterprise has sufficient volume of data backing it.


Q: Enterprises are certainly looking for ways to simplify implementation of AI-powered solutions. How has the learning at Nielsen motivated you to start Blackstraw?  

A: Nielsen is a phenomenal company that leads the space in which it operates. By nature of the business it operates in, Nielsen has use cases that need prediction and it is blessed with structured data. As a leader of the Global Innovations team at Nielsen, I led some of the trickiest industry challenges using AI. More importantly, the experience has taught me that all the challenges can be overcome if there is solid understanding of the process, intent, and focus to do it. I hope to use the experience I had at Nielsen towards a broader industry need.

Q: You said your tagline is “Simplify AI”. What are the top problems that mainstream enterprises get paralyzed by while formulating their AI strategy?

A: The three areas that are most problematic today in my view are: the “problem identification” problem, a talent gap to solution implementation, and access to appropriate AI building blocks.

Q: Great, let’s break those down individually. From the outside looking in, it seems obvious that enterprises should only pursue use of AI technologies when there is a legitimate business case. Please elaborate.

A: The term AI is broadly used to mean many things. In my opinion, if you have a use case where learning from old data can be applied to new data, you have a good use case to implement AI. But, you need to look at the volume and structure of data and have enough ROI to support the spend on building AI processes.

Q: Data scientist careers are in great demand today. Does the talent gap exist due to a people shortage problem or a skills and training one?

A: I call this the Artificial Talent gap problem. The frameworks that exist today have complicated the art of tuning parameters for model creation to a point that just the “top of the pyramid” data scientists can use the tools effectively. This limitation can be greatly simplified so that a broader talent pool can use it.

Q: A variety of data modeling frameworks are available for machine learning. What challenges will enterprises encounter when deciding what building blocks to adopt?

A: This is something else I learned while working at Nielsen. There are literally no labelling or operational tools that can be easily adopted for commercial re-use. This is a common problem across domains. Before enterprises can plan AI implementation, businesses first have to tune and modify the tool kits. And only then can they begin to approach solutions for the business problem at hand.

Q: So, how does Blackstraw propose to address these enterprise challenges?

A: First, we start with helping enterprises with use case identification; identifying situations where prediction is needed. This includes critically evaluating the state of data in enterprise and then solidifying the ROI and KPIs for the use case. Second, we are building a platform that provides smart AI building blocks for the enterprise’s domain and use case. Specifically, the building blocks span labelling tool kits, a modeling module that is abstracted, and an evaluation and deployment module. Ultimately, businesses need to operationalize learning within the organization, which includes quality control, baseline creation, and acceptable KPI management.

Q: You identified one of the mainstream business challenges for AI is overcoming the one-size-fits-all implementation approach. By commercializing a software platform, are you not restricting your market opportunity to those organizations that fit your model and approach?

A: We at Blackstraw realize that custom data and smarts built within the client ecosystems need easy integration to an AI toolkit for maximizing and expediting ROI. Our platform is designed to integrate into the client’s ecosystem. Our unique selling point is an end-to-end platform for smart AI implementation. From labelling to operationalization, we provide an integrated stack for AI adoption that can be used across enterprises and domains.

Q: Companies keeping control of the build vs. buy decisions is important. Can you share where Blackstraw is at in terms of financing the company and commercializing its platform and model?

A: Sure, Blackstraw is currently self-funded and completely focused on platform development. We have connected partners that we work with to provide services to clients. We are engaging with a handful of companies to validate our model. We expect to learn from these early adopters and iterate quickly to a world class platform.

Q: Best of luck! An innumerable number of enterprises are undertaking this AI journey. What advice can you provide them?

A: While AI seems like a relatively new topic in boardrooms and trade publications, there are a lot of individuals, companies, and academic researchers that have several years of hands-on experimentation and business application experience. These resources can ease your company into deploying the benefits of machine learning. You don’t have to pursue AI alone.  

Learn more at Blackstraw.ai.