Behind the Brains: Meet Dr. Daniel Rasmussen

Introducing Dr. Daniel Rasmussen. With a Ph.D. in computer science from the University of Waterloo and a postdoc at Princeton University, Dan's research is focused on building adaptable computational systems that draw from fields such as computational neuroscience, reinforcement learning, and deep learning. His unique background combines these (and other) approaches to create products that translate research advances into real world applications. Dan developed the first neural model capable of performing hierarchical reinforcement learning, as well as the NengoDL package for neuromorphic deep learning. At ABR, he is committed to pushing the boundaries of his research and turning academic projects into practical, cutting-edge solutions, as he leads the development of our NengoEdge software.

What inspired you to pursue a career in your field, and what led you to co-found ABR?‍

Having graduated from the University of Waterloo with a PhD in computer science, with a focus on theoretical neuroscience, I embarked on my professional journey in the realms of cognitive science and artificial intelligence. My initial experience as a research associate in Dr. Chris Eliasmith's lab at the Centre for Theoretical Neuroscience in Waterloo laid the groundwork for my future endeavours. In 2014, I co-founded ABR and pursued a postdoctoral fellowship at Princeton University. During this period, I delved deeper into the fascinating world of deep learning. In 2016, I went back to work for ABR after my postdoc, where I took the lead on the deep learning application development side of things, which grew into what we do now.

How would you describe your role at ABR, and what are some of your main responsibilities?

As the leader of our software team, I am responsible for the software side of product development at ABR. Our dedicated hardware team focuses on creating cutting-edge chips, while our software team ensures seamless communication and control over these chips.

“Our primary product, NengoEdge, is an advanced cloud-based tool designed for a wide range of audio machine learning applications. Closely integrated with our chip development, NengoEdge offers users an intuitive interface to interact with our chips, facilitating efficient implementation of machine learning models using their own data.”

What do you think sets ABR apart from other companies in the field of AI, and why do you enjoy working here?

At the core of our company's unique approach within the AI domain is our emphasis on optimizing edge hardware integration. Our expertise lies in deploying AI models in low-power environments with limited computational resources, significantly distinguishing us from our competitors.

“Also, our work with Legendre Memory Units (LMUs), a specialized recurrent neural network, allows us to achieve remarkable feats, such as reducing the power consumption of deep neural networks while simultaneously enhancing performance. Those two aspects combined are what sets us apart from other companies in the field of AI.”

I enjoy working at ABR because of the people here. We all came out of Chris's lab together and have been working together for quite a long time. We get along well, and we function well as a team. We get a lot of creative control over what we work on and the ability to direct our efforts as we think best. And from a technical perspective, the problems that we work on are fascinating and keep me excited to come to work.

How has working at ABR impacted your professional growth and development, and what skills have you gained?

I've developed my specialty in machine learning and deep learning over my time at ABR. While my postdoctoral research provided me with a solid theoretical foundation, my six years of hands-on experience have allowed me to transform these theories into practical applications for real-world products. I have gained skills in implementing complex software projects, leading a team through such endeavours, and making all those pieces work together to create a successful product. This accumulation of experience and knowledge has been the primary growth factor in my professional journey.

Besides sharpening my technical skills, I've also picked up some handy business know-how during my time at ABR, even though it was outside my original skill set. I've learned about the many sides of running a business, from hiring talented people to managing a team so we can work together and reach our common goals. This mix of skills has made me a more well-rounded professional and better prepared to help us succeed as a company.‍

Can you share any exciting developments or upcoming projects that you are currently working on?

“The main thing we're working on is preparing to launch our dev kits equipped with our custom-designed chips. As software team members, we focus on crafting an extensive suite of tools that facilitate seamless interaction with the chip, allowing users to run their models and achieve optimal results.”

It will be a big push, but we're excited to create something that will be so central to our products, and for people to get out there and use it. It's exciting for the whole team.

What currently has you the most excited about the future of AI?

AI keeps improving every year, with new powerful models that lead to even better applications. It's becoming a regular part of our daily lives, giving us real benefits like identifying objects in our photos or answering questions with ChatGPT. AI lets us do amazing things we couldn't do before, which will only speed up in the future. I bet when my kids grow up, ChatGPT will be as everyday as smartphones are for us. I'm super excited about all the possibilities AI's evolution will bring.

What role do you see ABR playing in the future?

“Our strength at ABR is bringing models like ChatGPT to life by integrating them into edge devices, allowing people to interact with AI in everyday routines. We aim to make AI a tangible part of people's lives, not just a concept that resides in remote supercomputers.”

The products we create work towards this goal, and I find that pretty exciting. I envision voice interfaces becoming increasingly prevalent in the future, replacing many button-based interactions. We're currently in a transitional phase where voice interfaces haven't quite surpassed buttons in terms of usability. However, once we overcome this obstacle and develop truly efficient voice interfaces, I believe they will become the primary mode of interaction with various devices, including cars, kitchen appliances, and other household items.