Dr. Sam Zheng, CEO & Co-Founder of DeepHow – Interview Series

Sam Zheng, CEO and Co-Founder of DeepHow, spearheads a rapidly evolving startup, backed by esteemed investors. DeepHow revolutionizes skilled workforce training with an innovative, AI-powered, video-centric knowledge capturing and transfer platform.

Prior to DeepHow, Sam dedicated over a decade to Siemens, driving digital innovation across various industries. His noteworthy projects, such as the Cloud Digital Inspection Jacket, have significantly improved technical knowledge sharing, efficiency, and user experience, earning his team the prestigious Siemens Innovation Award.

Simultaneously, Sam serves as an Adjunct Professor of Psychology at Tsinghua University and holds a Ph.D. in Engineering Psychology and a Master’s in Statistics from the University of Illinois at Urbana-Champaign.

You have an educational background in psychology and statistics, how did you transition to focusing on videos and machine learning?

My background in psychology and statistics actually served as a natural segue into the realm of machine learning and video-centric platforms. The study of psychology sparked my fascination with the human mind and intelligence, particularly the process of skills learning and expertise development. Meanwhile, statistics provided the mathematical foundation to explore artificial neural networks, inspired by our biological brain.

In today's digital age, videos have emerged as a more engaging, interactive, and effective learning medium. This shift is evident with platforms like YouTube and TikTok, where users, particularly the younger generation, invest hours consuming and learning from video content.

However, the traditional process of creating instructional or how-to videos, especially the editing part, is time-consuming and labor-intensive. A short video spanning a few minutes could require hours of meticulous work. Recognizing this inefficiency and the potential to enhance the learning experience, my co-founders and I decided to harness the power of artificial intelligence (AI) and machine learning to expedite the how-to video creation process.

Our AI-powered video platform can transform hours of laborious work into mere minutes, drastically improving efficiency without compromising effectiveness. In essence, my academic grounding in understanding human cognition and the statistical models that mimic it helped pave the way for this innovative venture.      

You have multiple patents under your name, what’s the most important one that you worked on?

All my patents focus on harnessing technology to enhance human performance. During my tenure at Siemens, one notable project involved working on a solution for Professor Stephen Hawking. We developed an intuitive eye-typing input method to aid individuals with ALS, similar to Professor Hawking's condition. This innovative work is now part of a pending patent.

However, the most significant patent I've contributed to is a recent one: the Generative AI-Powered Know-How Management Platform for Industrial and Manufacturing Organizations.

Here's a brief overview:

Our invention presents a cutting-edge generative AI solution specifically tailored for industrial and manufacturing organizations. It effectively captures, organizes, and disseminates both institutional and tribal knowledge, intersecting fields like artificial intelligence, machine learning, training and development, and knowledge management systems.

Industrial and manufacturing sectors often face immense challenges in capturing, organizing, and sharing critical knowledge. High staff turnover rates, complex processes, and the perpetual need for upskilling amplify these difficulties. Traditional knowledge management methods are often cumbersome, time-intensive, and lack flexibility, necessitating a more advanced solution.

Our generative AI solution employs proprietary algorithms and machine learning techniques to streamline the creation of video-based standard operating procedures (SOPs), optimize workflows, and facilitate quick, efficient access to information via AI-driven chat features. Due to its adaptability and scalability, our solution is suited to a wide range of manufacturing contexts.

Could you share the genesis story behind DeepHow?

Before we launched DeepHow, I was working alongside our other founders, Patrik Matos da Silva and Wei-Liang Kao, at Siemens, driving a variety of digital innovation projects in industrial and manufacturing sectors. Our journey took a big leap when we joined the Techstars Mobility Accelerator in Detroit in 2018. The experience was essentially a boot camp, helping us validate our idea, identify gaps, and connect with potential investors, partners, and mentors.

We saw the value in skills and experiences that people had cultivated over the years but there was one problem- there wasn’t a truly effective way to capture and share this expertise. We observed how rapidly technology was advancing, and it occurred to us that the methods we used for employee training hadn't kept up. We were still relying on outdated, time-consuming methods that weren't efficient and, frankly, weren’t engaging enough.

I realized there was a unique opportunity to combine advances in AI and video technology to completely change how we capture, structure, and share information. So we set out to build DeepHow, a platform where companies can create amazing training videos in-house and remain in control. To actualize this idea, we created an AI system called “Stephanie”. Stephanie is the backbone of our solution, capturing experts' know-how and creating a knowledge repository. It generates step-by-step, interactive how-to videos to speed up the learning process for new and less experienced employees.

Through the use of AI workflow indexing and segmentation, we've managed to create video content ten times faster and improve workforce performance by 25%.

We’ve come a long way since the early days, but our mission is still the same. We are committed to helping companies leverage their wealth of institutional knowledge and expertise and enable their employees to continually learn and grow. It’s a thrilling journey.

What are some of the workplace challenges DeepHow is solving?

Knowledge Transfer: In any workplace, transferring skills and knowledge, especially from seasoned employees to new hires or less experienced staff, can be quite a challenge. We have made it possible to capture and structure that knowledge in a way that's simple to share, broken down into segments and easier to grasp.

Training Efficiency: Traditional training can be tedious, right? Long hours, hard to retain information- it's not always the most efficient process. With DeepHow, employees can learn at their own speed and in a way that is better tailored to their own unique learning style. It’s our goal to make training painless and enjoyable.

Skills Gap: Sometimes, figuring out where your teams' skills gaps are can feel like trying to find a needle in a haystack. That’s another thing that we can help with. Our platform uses analytics to show you where there is a gap in training and enables training teams to bridge that by creating the specific content they need.

Outdated Training Materials: Things change fast, don’t they? Standards and procedures, they’re all constantly evolving. Well, we make sure that your training materials are never left behind. Editing and updating is quick and painless, and being able to share those updates across different locations and shirts with a simple skills assignment makes sure that no one is left without the most up-to-date information.

On-Demand Learning: Convenience is king! Everyone wants to have access to what they need, when they need it. That’s the philosophy behind DeepHow. We believe that training shouldn't be a scheduled event that disrupts the entire day, but a flexible resource that’s there when you need it. Being confined by a place and time isn’t ideal. Need to learn something specific, right now? Search and watch. It couldn’t be easier.

Boosting Employee Engagement: We all know that training can sometimes be, well, not the most exciting thing. But it absolutely does not have to be this way. DeepHow allows creators room to be creative. Learning should be a fun, engaging and enjoyable experience.

Fostering Communication: Communication can be tricky, especially when dealing with a complex process or procedure. Our platform makes communication simpler by enabling step-by-step guides that allow employees to easily digest and understand the task at hand, promoting clear and consistent messaging across the board. Our platform also understands, translates, and transcribes in nearly 50 languages and counting. This feature alone has proven to be one of the most valuable tools many companies possess. Allowing someone to learn in their native language ensures better understanding and boosts morale.

How does DeepHow enable enterprises to create an adaptive training program?

Let’s consider the traditional training environment. You’ve got static material, rigid schedules, and a one-size-fits-all approach. Now, these methods don’t take into account that everyone learns differently and at a different pace. They aren't scalable or flexible to accommodate the rapidly changing landscape or individual employee progress. Those are significant pain points for any enterprise, right?

That’s exactly where DeepHow steps in. We help you flip the script on these issues. We enable businesses to develop agile training programs that are dynamic, personalized, and highly responsive to business needs and employee learning patterns. 

Our platform captures expert knowledge in easy-to-follow, video-based learning modules. But we don’t stop there. We harness the power of AI to analyze employees' interactions with these models, enabling deeper insight into where your skills gaps remain. It’s about turning weaknesses into strengths and capitalizing on individual areas of expertise to foster a continuous learning culture. 

What role does DeepHow play in increasing workplace safety?

Safety is such a critical aspect of any workplace, yet it is often difficult to get it right, leading to unfortunate accidents and regulations violations. This is unfortunately true in industries like manufacturing, construction, or healthcare, where even a small lapse can have significant consequences.

So where does DeepHow fit into this picture? Well, we are passionate about ensuring that safety practices are clearly understood and consistently implemented across the board.

We do this by providing a platform where companies can easily capture and share expert knowledge on safety protocols. Instead of the old-school, hard-to-follow manuals, we offer interactive, step-by-step video guides. They’re straightforward, easy to understand, and most importantly, accessible anytime, anywhere. This means employees never have an excuse to not stay up to date with the latest safety protocols. Safety should always be a top priority and everyone deserves to feel safe at work.

How is Generative AI used to enable this to happen?

At the heart of our Generative AI-powered know-how management platform, “Maven,” lies a suite of innovative capabilities aimed at revolutionizing the way industrial and manufacturing organizations capture, manage, and share their institutional and tribal knowledge. Maven employs proprietary algorithms and advanced machine learning techniques to simplify and enhance a variety of processes:

1. Streamlined Video SOP Creation: Using our unique Generative AI algorithms, Maven assists in the automatic generation of video shooting guides and narration scripts. This facilitates the creation of high-quality video SOPs with consistency, setting a new industry standard.
2. Efficient Workflow Optimization: By leveraging advanced machine learning techniques, Maven's workflow visualization tools enable users to simplify complex tasks, thus promoting productivity through user-friendly interfaces and smart applications of AI.
3. Strategic Knowledge Mapping: Maven's AI algorithms help organizations map crucial know-how, identify knowledge gaps, and uncover essential content opportunities. This results in the development of targeted training materials uniquely tailored to the needs of a skilled workforce.
4. AI-Enhanced Chat Feature: Powered by leading-edge natural language processing algorithms, Maven's AI Chat enables users to access necessary information quickly and efficiently using natural language queries.
5. Multimodal Content Generation: In situations where a written SOP is unavailable, Maven can analyze video content and generate multimodal content like step-by-step video SOPs, text SOPs, and workflow diagrams, thereby streamlining the content creation process.

By marrying advanced Generative AI with the latest knowledge management strategies, Maven offers organizations a unique tool to harness their skilled workforce's potential, driving continuous improvement and innovation.

Are there other types of machine learning algorithms that are used?

Indeed, DeepHow leverages a range of sophisticated machine learning algorithms and AI techniques within the realms of Natural Language Processing (NLP) and Computer Vision. These techniques, both supervised and unsupervised, underpin our proprietary, domain-specific AI technology, which has been trained and optimized for the industrial and manufacturing sector. Key areas of application include:

1) Workflow Segmentation: We utilize machine learning algorithms to extract critical information and steps from complex, unstructured task demonstrations captured in videos. This enables us to break down complex procedures into manageable, teachable steps.

2) Multimodal Step Embedding: By modeling an ‘activity genome', we're able to reconfigure instruction and workflow guidance to better suit specific operational contexts.

3) Cross-Modality Retrieval: We use advanced in-video search techniques to facilitate multi-language, skills-based content retrieval. This allows users to access relevant information with greater efficiency and precision.

4) Know-How Mapping: We construct a knowledge graph that visually represents an organization's core competencies. This mapping allows companies to identify their knowledge assets clearly, enabling more effective upskilling and training strategies.

These advanced machine learning techniques, combined with our focus on industrial and manufacturing services, allow us to offer a comprehensive solution to the unique challenges faced by organizations in these sectors.

For companies that wish to get started, what’s the process?

We’ve designed our platform with simplicity in mind, so onboarding your company doesn’t have to be complex. In fact, more than 80+ enterprise and SMB customers have successfully deployed our solutions to over 400 sites in 24 countries across 6 continents. 

First, our teams will meet and have a conversation about your company’s specific needs and pain points. We want to understand your goals, the training challenges you’re facing, the kind of skills your workforce needs – the whole picture. 

Next, we guide you through the process of capturing your experts' knowledge. This can be about any process or skill that is important to your company. Our team will assist you in creating these step-by-step video guides using DeepHow’s intuitive tools.

Our team is there to support you every step of the way, from the initial setup to the ongoing optimization of your training program. We’re here to form a partnership with you and enhance your workforce's skills and efficiency. Simply visit DeepHow.com to get started.

Is there anything else that you would like to share about DeepHow?

At the heart of DeepHow is a clear, compelling mission: we aim to empower every skilled worker to become an expert. We strive to make knowledge transfer and training seamless, engaging, and cost-effective, harnessing the transformative power of AI. We firmly believe that technology should augment human capabilities, not replace them. This principle guides everything we do.

In our rapidly evolving technological landscape, this mission is more relevant than ever. The shift towards digital transformation and Industry 4.0 has modernized manufacturing, introducing an array of advanced technologies. These innovations have significantly altered job requirements, demanding workers to acquire new technical skills to operate, maintain, and optimize these sophisticated machines. The pace of change is so rapid that traditional training approaches struggle to keep up, leading to an ever-widening skills gap.

Our aim is to address this challenge head-on, empowering workers to ‘build back better' by reskilling for the factories of tomorrow. High levels of automation mean there's less need for manual labor; instead, the focus shifts to leveraging workers' expertise and intuition in operating advanced technological systems.

Factories have evolved substantially over the past decade, with the integration of robotics, cobotics, and analytic technologies that continually optimize output and minimize waste. To manage these technologies, an upskilled workforce is critical.

DeepHow provides a modern training methodology, enabling manufacturers to attract talent, fill entry-level positions, and progressively upskill workers for advanced roles across manufacturing, logistics, and planning. With an emphasis on contemporary, engaging training, we're helping to shift the perception of manufacturing from a dead-end career to a dynamic, technology-driven field with limitless possibilities.

Thank you for the great interview, readers who wish to learn more should visit DeepHow.