Dr. Nikunj Mehta of Falkonry
An exclusive Tech Tribune Q&A with Dr. Nikunj Mehta (founder and CEO) of Falkonry, which was honored in our:
Tell us the origin story of Falkonry – what problem were you trying to solve and why?
Falkonry began with the goal of enabling fact-based, data-driven decision-making for industrial companies and national security organizations. We set out trying to solve the challenge of analyzing the vast amount of time series data produced by machines and sensors so that organizations could use it effectively to improve their operations. Back in the day when these measurements were recorded by hand, it was possible to interpret that data manually or with statistical or physics-based analytics tools. As industries seek to substantially improve operations with data, they are producing way too much data for conventional methods. To overcome this challenge, Falkonry invented an AI to discover patterns in this data and bring specific events to the attention of human operators, thereby enabling a more nuanced and data-driven understanding of operations, and allowing ops teams to take timely actions for further improvements.
What was the biggest hurdle you encountered in your journey?
Early on we realized there needs to be a way to combine AI into end-user applications. How can we rapidly offer solutions that are infused with AI and are directly usable by end-users in our customer organizations? – that was the question we were trying to answer. You see the data that these organizations generate are a lot like the ECG reports you encounter at your cardiologist. While specialists can provide reliable review of such data, end users find themselves getting overwhelmed with data to review and often need a second opinion. The main challenge is to automatically identify events from such data without asking subject matter experts to do it manually, and the hurdle to overcome is to find just the right events so as not to overwhelm experts. The second big challenge was building trust: How do we develop confidence in our customers’ minds that the insights that the AI is surfacing are actually the result of some physical phenomenon and not some spurious correlation? Trust requires providing visibility into the internal workings in a simple manner so that stakeholders can correctly interpret findings without being overwhelmed.
The way to do that was to provide an explanation for the AI’s findings in ways that they can understand. We wanted to put our insights directly in the hands of an end-user (in most cases, a reliability engineer or a production engineer). Their way of looking at the world is very much focused on the root causes of failures, the effects of those failures in the data, and consequent outcomes on production. We had an intuition about this at the very beginning of our journey – if we can overcome these two challenges in a general-purpose manner, we would be able to create a higher value addition through software with only a thin veneer of services. That flipping of proportions is the basis of the Falkonry business and our unique value proposition.
What does the future hold for Falkonry?
Our mission is to make the physical world’s information accessible and usable. Utilizing the power of GPUs, backed by our recent partnership with NVIDIA, our AI has the potential to deploy at petascale and in a novel edge to cloud architecture. This kind of never-seen-before speed and scalability will allow us to take on several audacious challenges: autonomy in transportation, zero defects in manufacturing, carbon neutrality in industrial activity, and zero surprises in healthcare. Expanding our sphere of operation, Falkonry is on its way to not only revolutionizing smart manufacturing at scale, but also bringing its AI-driven analysis platform to new domains such as cybersecurity, IT systems, and vision systems. The future is bright for smart decision-making, and Falkonry’s time series AI suite is here to accelerate this transformation.
What are your thoughts on the local tech startup scene in Cupertino?
I have lived in Cupertino for the last 18 years and it has been my dream to design in Cupertino alongside Apple and Amazon Labs. Because of high educational attainment, the people of Cupertino generally work in the best Silicon Valley companies. There have also been many successful Cupertino startups such as Bromium, BlueKai, Rancher Labs, Demisto, and Cloud.com that were acquired by various large companies like Palo Alto Networks, HPE, and Oracle. Cupertino startups are often known to focus on software for businesses, owing to its talent pool and the heavy expertise its people have developed working in that sector for the last twenty years. Therefore, when we found the right space to build Falkonry further in Cupertino, we immediately jumped on it last year to move here.
What’s your best advice for aspiring entrepreneurs?
Since my expertise lies in artificial intelligence, I can offer advice to entrepreneurs who are answering that particular calling. Opportunities to apply AI exist virtually in any industry that hires analysts to look at trends and data. The computing, scaling, and core AI technologies today are mature enough for broad application to different types of data. Aspiring entrepreneurs should offer vertically integrated business software around that AI to create efficiencies in the hands of their users. I would encourage entrepreneurs who are just starting out to think about the entire flow of intelligence from the AI, right down to the intended user who will be making decisions based on AI outputs, all the way to how that produces an impact on the people that make up that organization.
Understanding how they can improve the nuances of that interaction and the lifecycle of that intelligence is the crux. Technologists and AI innovators often fall into the trap of concentrating just on the core technology, but they mustn’t lose sight of the human involvement in the systems they are designing. Strategic ROI benefits of AI come from lifting the entire organization to the next level of capability and therefore, the buy-in of end-users whose work is being supplemented or enhanced is very important. Moreover, along with surfacing AI insights, the system should be able to capture data about the actions taken based on the insights and their outcomes, thereby facilitating continuous learning.