AiThority Interview with Yashar Behzadi, CEO and Founder at Synthesis AI

Hi, Yashar. Please tell us about your journey into the AI and Data industry?

My journey into AI and Data started in graduate school, where I earned a Ph.D. from UCSD in 2006 focused on spatial & temporal modeling of brain imaging data. I then worked in Silicon Valley at the intersection of sensors, data, and machine learning across industries for the next 12 years. I was fortunate to work on some amazing technologies and have over 30 patents issued or filed focused on signal processing, machine learning and data science.

Before founding Synthesis AI in 2019, I led a global AI services company focused on developing computer vision models for leading technology enterprises. Regardless of the company’s size, every engagement was limited by the amount and quality of labeled training data. As companies developed new hardware, new models or expanded their geographic and customer base, new training data was required to ensure models performed adequately. More fundamentally, it became apparent that many future product features and models could simply not be built with today’s human-in-the-loop annotation paradigm. Humans simply could not label key attributes related to 3D position, object interactions, material properties, lighting, etc. Companies were also struggling with ethical issues related to model bias and consumer privacy, which were difficult to deal with current model development processes.

I established Synthesis AI to transform the computer vision paradigm. The company’s synthetic data-generation platform enables on-demand generation of photorealistic image data with an expanded set of pixel-perfect labels. Our mission is to pioneer synthetic data technologies to enable the ethical development of more capable models.

How did you cope with the disruptions brought by the pandemic? How did you leverage technology to stay on top of your product development roadmap during the peak of the pandemic months?

The pandemic is a once-in-a-lifetime event, and it has disrupted how we live, work and interact. As a start-up, we adapted very quickly and implemented a distributed and remote working model. For us, it allowed us to scale faster and reach talent pools beyond Silicon Valley. We leverage several tools, but more importantly, we cultivated a culture of communication and collaboration within the team. Each team member gained more autonomy, we all embraced a culture of accountability.

For our particular business, the pandemic was a catalyst. As companies were no longer able to capture real-world data to power their AI systems, the ability to leverage synthetic data became a strategic advantage.

Recommended: AiThority Interview with Anthony Scriffignano, Ph.D., SVP and Chief Data Scientist at Dun & Bradstreet

What is the most contemporary definition of Synthetic Data? How do you promote this idea across your customer base?

Synthetic data is computer-generated image data that models the real world. Technologies from the visual effects industry are coupled with generative neural networks to create vast amounts of photorealistic and automatically labeled image data. Synthetic data allows for creating training data at a fraction of the cost and time of current approaches. In addition, since the data is generated, there are no underlying privacy concerns. Synthetic data enables the efficient prototyping, building, and testing of complex computer vision systems.

Currently, it’s common for most AI systems to leverage ‘supervised learning’ or the process in which humans label and effectively teach AI how to interpret images. This process is time- and resource-intensive, in addition to being limited to what humans can effectively label. Historically, there has also been little regard for consumer privacy, and these concerns are amplified as systems increasingly use human facial data in our day-to-day environments. Our approach is to create photorealistic digital worlds in which complex image data can be synthesized. Since we generate the data, we know everything about the scenes, including never before available information about the 3D location of objects and their complex interactions with one another and the environment. We are able to create millions of perfectly labeled images on demand. Acquiring and labeling this amount of data using current approaches would take months, if not years. This new paradigm will enable a 100x improvement in efficiency and cost and drive a new class of more capable models.

We are in the process of developing a report that highlights the discrepancies in how practitioners define synthetic data, as well as perceived barriers to further adoption. You’ll be able to access the report on the Synthesis AI website in a few weeks.

What kind of IT infrastructure does a company require to better leverage your Data Management platform? How do you help your customers achieve this?

Machine learning requires a ‘data-centric’ infrastructure and ML developer operations to support it. We enable companies by supplying the cloud infrastructure and creating simple API interfaces to enable the generation of synthetic data. By using an API-centric approach, we enable machine learning engineers to focus on the data and not the infrastructure.

Could you tell us more about Embedded Analytics and how AI analytics technology investments play a role in the adoption of these applications across industries?

Ultimately data, whether real or synthetic, is used to train AI models. The AI models are then deployed in the cloud or to edge devices in embedded systems. Computer vision is leveraged in the devices in our pockets, homes, businesses, cars, factories, and more. We believe synthetic data will enable the next generation of computer vision and AI. Synthetic data is orders of magnitude faster and cheaper than traditional human-annotated real-data approaches and will come to accelerate the deployment of new and more capable models across industries.

Recommended: AiThority Interview with Josh Harbert, Chief Marketing Officer at Delphix

What are the “Hits” and “Misses” with AI in the automotive industry? How do you use AI and ML in Synthesis AI?

For the safe and general deployment of Autonomous vehicles, AI systems need to perceive the world reliably and make proper decisions across a wide variety of situations. AI systems are only as good as the data they are fed, and AV companies have spent vast amounts of money and time building and deploying cars to collect diverse datasets. However, it is impractical, if not impossible, to capture sufficient examples of rare events such as accidents, unusual weather, and unpredictable driver or pedestrian behavior. Further complicating data collection are privacy concerns around personal data collection, rapidly evolving sensors (cameras, lidar, radar) and hardware which often require recollection of data with each new generation of sensors.

AI and ML are at the core of Synthesis AI’s data generation platform. We leverage modern generative neural networks together with traditional cinematic visual effects pipelines to create photorealistic labeled data. This image data is used to train computer vision AI models across industries.

Tell us more about hiring trends in the AI/ML and Voice market influenced by AI and data science:

All industries will be impacted by AI/ML, and companies will need to build in-house capability to avoid being disrupted.  As with previous technology trends (e.g., internet, mobile, cloud), the companies that proactively hire and build capability will emerge as sector leaders.

Industry Trends & Insights

Which industries have you been keenly following to understand current technology trends?

In AI/ML, technology companies tend to be early adopters, and verticals such as mobile phones, cloud computing, autonomous vehicles, and robotics have aggressively embraced cutting-edge technologies. I keep informed of the latest industry and academic trends to ensure our offerings are ahead of the curve.

Where is the AI-based mobility industry heading? Which technologies are you keenly following in this industry?

Computer vision, speech recognition, and NLP will continue to advance at an increasingly rapid pace driving amazing new capabilities. Synthetic data, unsupervised learning, and reinforcement learning are key technologies to watch in the coming years.

Recommended: AiThority Interview with Dan O’Connell, Chief Strategy Officer at Dialpad

What’s your favorite podcast/webinar that you have listened to in recent times?

Two Minute Papers is a favorite of mine as it is very informative and presents the latest research in an enthusiastic and entertaining way.

Tag a person in the industry whose answers you would like to see here:

Andrew Rabinovich PhD, LinkedIn

Sergey Nikolenko PhD, LinkedIn

Thank you, Yashar! That was fun and we hope to see you back on AiThority.com soon.

[To share your insights with us, please write to sghosh@martechseries.com]