CTO Sessions: Anatoli Gorchet, Neurala

What type of CTO are you? "I am the type of CTO who likes to challenge the norm, coming up with my own version of a solution."


Name: Anatoli Gorchet

Company: Neurala

Job title: Co-founder and CTO

Date started current role: March 2006

Location: Boston, MA

Anatoli Gorchet is the co-founder and CTO of vision AI company Neurala. Gorchet has over 20 years of experience developing massively parallel software for neural computation, he is a pioneer in applying general-purpose computing on graphics processing units to neural modeling. Gorchet has spoken at every major neural network conference as well as at GTC, DARPA, The National Institute for Aerospace and a keynote at the Embedded Systems Conference. He holds several patents, has authored over 30 publications on neural networks, and advises Fortune 500 companies on how to use AI to improve operational efficiencies.

What was your first job? My first job was what we called “industrial climbing” – you know those guys who hang on ropes outside of your office and wash your windows, paint the frames, or fix insulation? That was it.

Did you always want to work in IT? If you look at the technical definition of IT, it is “the study or use of systems (especially computers and telecommunications) for storing, retrieving, and sending information.” I never really wanted to work in it, and never will. I am not passionate about the world of storing, sending, and retrieving – I prefer the analysing and manipulating of data and information.

What was your education? Do you hold any certifications? What are they? I started at St. Petersburg Polytechnic University (former Leningrad Polytechnic Institute). During my senior year I came via an exchange program to Belmont University and received my BS in computer Science there in 1997. Then I graduated from Middle Tennessee State University (MTSU) in 2000 with an MS in Computer Science. After that, I got my Ph.D. in Cognitive and Neural Systems from Boston University.

Explain your career path. Did you take any detours? If so, discuss. I would say it took me a while with a few detours, but I did finally end up where I’m supposed to be. While at Belmont University I discovered the field of AI and have been studying intelligence, both artificial and natural, ever since. I set up my master’s at MTSU with double minor in psychology and biology, because from the start I believed that making AI without knowing how nature does things is nonsense. The brain is deeply parallel, so I concentrated on parallel computer architectures and algorithms, then wrote a parallel neural simulator for my master’s thesis.

During my last months at MTSU I was invited to work as a Lead AI Developer for Surfari, Inc. It was a temporary gig as I was already set on starting my Ph.D. at Boston University, but it gave me a feel of how life is in the software industry. I continued to develop my neural simulator and used it for my Ph.D. research. I also started to share it with my fellow Ph.D. candidates, ultimately impressing two of those fellow candidates – Max Versace and Heather Ames – which led to them suggesting we start the company that would become Neurala.

Through our subsequent research, we created our now patented Lifelong-DNN (L-DNN) technology which dramatically reduces the data requirements for AI model development and enables continuous learning in the cloud or at the edge. Meanwhile, I continue to lead the development of our technology as CTO.

What type of CTO are you? I am the type of CTO who likes to challenge the norm, coming up with my own version of a solution. The only problem is that stops being needed after a Series A funding round. So, I am trying to transition to become a CTO that is a product driven researcher and organiser.

Which emerging technology are you most excited about the prospect of? One trend I’m excited about is advancements in deep learning technology to enable AI to learn continuously over time. I think this has a lot of potential in industries such as manufacturing, where the production line is constantly changing as new products are added and removed. Traditional approaches to training DNNs require large amounts of data, computer power and time. They also need to be ‘reprogrammed’ anytime a new variable is introduced. But with L-DNN, AI is able to learn quickly, at the edge/on-device, and continuously over time. This means that rather than having to retrain AI systems for new use cases or situations that are unfamiliar, AI is able to keep up and adjust on the fly.

Are there any technologies which you think are overhyped? Why? Ironically, one of the most hyped technologies right now is AI. Everyone says they do AI, but estimates have shown that there are only about  300,000 AI experts – 22k of which have PhDs – in the world. So while there are a lot of people and companies out there who claim to do AI, I wouldn’t be so sure. I would also caution anyone looking to implement AI to make sure they know what they are getting into, as there are a lot of myths and misconceptions about how to successfully and effectively develop AI.

What is one unique initiative that you’ve employed over the last 12 months that you’re really proud of? This year, we launched our VIA (Vision Inspection Automation) software: AI to help manufacturers improve quality inspection on the production line while adapting to fluctuations in consumer demand. Historically, AI has been too expensive or complex to deploy at scale in a manufacturing environment. But manufacturers today are in need of automation and AI more than ever before, as the industry adjusts to a new normal: heightened pressures on machine utilization; production efficiencies and quality control; massive amounts of data from industrial IoT systems; and the need to manage all of this with fewer people on the factory floor.

VIA enables manufacturers who have not worked with AI before to easily train and implement vision AI as a part of their workflow. With less data required and faster training, VIA improves return on investment by automating the quality inspection processes that were not previously viable – removing the need for human intervention and achieving 100% inspection rate. And, with the ability to run directly on existing hardware on the factory floor, VIA makes AI accessible to industrial automation users who prefer not to rely on internet access or connectivity to the cloud. As a result, manufacturers can keep their data on the factory site, without concerns about privacy or lag time that are typically associated with cloud deployments.

Are you leading a digital transformation? If so, does it emphasise customer experience and revenue growth or operational efficiency? If both, how do you balance the two? We are not in the middle of our own digital transformation, but we are helping our customers in the manufacturing space adopt automation and implement Industry 4.0 initiatives.

What is the biggest issue that you’re helping customers with at the moment? Today’s factory floors are more complex than ever, with IIoT systems collecting massive amounts of product data, as well as basic diagnostics from industrial equipment. We are helping customers such as global manufacturing leader IMA overcome this complexity by delivering AI solutions for industrial machines. As the IIoT becomes commonplace, manufacturers need human-level AI that can extract actionable insights from the data at the compute edge. We are working with IMA to deliver our AI directly on their industrial machines, at the compute edge, so that operators can quickly and independently set up advanced AI systems without requiring specialised expertise. 

How do you align your technology use to meet business goals? Our goal is to make AI more accessible for manufacturers. We’ve accomplished this by making our technology extremely adaptable and easy to use. We don’t think you need to be an AI expert to use it, so we’ve designed our VIA software accordingly.

Do you have any trouble matching product/service strategy with tech strategy? I do. Product strategy requires stability and gradual improvement – you don’t fix things that aren’t broken. Meanwhile, true tech advancement requires throwing away pieces that appear working and replacing them with pieces that might one day work better.

What makes an effective tech strategy? You need a vision and the confidence to challenge the “expert’s opinion” on the technology. For example, we looked at traditional approached to DNN, and thought about how they could be improved. From there, you need to establish a team that will both support your vision and question you when it makes sense. Together, you will be able to align on a strategy to help push your technology out to the masses.

What predictions do you have for the role of the CTO in the future? Technology is always changing, so as CTO, I think you need to be adaptable and ready to pivot your product to match the times. In the future, we’ll see more of this as we continue to live in this “new normal.”

What has been your greatest career achievement? Throughout my whole life I have never had a job that I did not enjoy one way or the other. Whether you are making a billion dollars or winning a Nobel Prize, it doesn’t mean anything if you are not having fun and passionate about what you are doing.

Looking back with 20:20 hindsight, what would you have done differently? Back in 2004 I made a mistake during my postdoc position interview. I would have fixed that, and then I wouldn’t be a CTO at this stage in my life. Being CTO is quite enjoyable, but could have been more so when I was younger and had less responsibilities.

What are you reading now? This morning I was comparing levels of purity in English (about 1590AD) to Russian (Soviet era) translations of Metamorphoses of Apuleius.

Most people don't know that I… am an open book and will tell you anything about me – all you have to do is ask!

In my spare time, I like to…Dive deep into some many hundreds of hours-long fantasy computer game or book series and forget that this world exists.

Ask me to do anything but… take life seriously.