Outstanding medical products require passionate makers: Behind our unique 3D mobility app is an interdisciplinary team with more than 40 brains. In our blog series People@Lindera, we introduce the various experts: Today we dive into the world of data scientist Bowen Zhou. Before joining Lindera’s mission to make precision and preventive medicine accessible to all and to push the boundaries of human movement, she did research in robotics at Karlsruhe Institute of Technology. In the following article, she tells us why she ditched her research assistantship at the university for the Berlin-based medtech company and is now driving the data-driven transformation of our healthcare:
Bowen, you are a data scientist. What exactly do you do?
It is my job to apply machine learning techniques and computer vision. algorithms to 2d/3d pose estimation. I take care of the precision and reliability of our digital mobility analysis, which is based on artificial intelligence. More specifically, I make sure that our estimation of the human pose is more accurate.
Who do you get your data sets from – what data do you process?
When I tell people about my job, many first think that I record people walking all day to collect data. Thank God, that’s not the case. For human pose estimation, there are already widely used open source datasets, such as the Microsoft Common Objects in Context (COCO) dataset for 2D and Human3.6M, the most widely used indoor dataset for 3D HPE from monocular images and videos. Our team has been working with Human3.6M to train a neural network. Currently, I have been working on a new training round using motion capture data. The advantage to such open source datasets is the structured nature of the data, amount of data available and data points included. Public datasets are open to all and contribute to a lot of research and development.
Data security is a sensitive issue in the health sector. How do you ensure that people’s sensitive data stays safe with you?
We take our data protection role very seriously and strictly adhere to the data security requirements for digital health applications to ensure the security of personal data.
What are you currently working on?
I am currently working on the 3D human pose estimation algorithm and 2D pose tracking pipeline for our LTech products. With our LTech Software Development Kit, fitness and health app providers can offer personalised training feedback to their users. When our technology is built into any health app, Deep Learning is used to generate real-time feedback easily through the app. There is no need for multi-camera systems or complex deep sensors – the smartphone or tablet camera is perfectly adequate. We have scored a real coup with this technology.
Why is it important for our health to be more data-driven?
Data-driven healthcare is our future, this is already emerging internationally and with numerous SDK and platform offerings – we saw this for ourselves at the HIMSS Global Health Conference & Exhibition. We have long lived in a data-driven world in our private lives. With the help of mobile phones, smartwatches and other wearables, we can collect – and further utilise – a multitude of information and data about our health and activities. The problem is: so far, we are still making too little use of it. Yet there is so much potential in our data for prevention and early detection of health anomalies.
For data scientists like me, the status quo is already a true data paradise. Due to the large number of mobile devices, I have the chance to analyse a lot of data to get a clearer idea of health and fitness needs and to draw concrete conclusions – for example, for the quality of senior care in nursing. With the results of my work, seniors can specifically improve their mobility by being educated about their posture, gait and fall risks. Simply by analysing a video of a person’s gait, for which I train the algorithms on the various gait parameters. By the way, we also work with a questionnaire.
Why is it important that all people are „data literate“?
Data literacy means nothing more than being able to read, use, analyse and question data. The focus of data literacy should not be on mathematics, algorithms or technology. Rather, it should relate to each and every one of us. It is obvious that data literacy is important for companies and businesses. After all, only a team that has sufficient data literacy is able to deal with different data sources, know what the data can tell us, and provide continuous insights and decision support. We can say that data literacy at Lindera has become more than a mathematical skill – it has become a life and work skill.
Which KPI is most meaningful to you and why?
That’s an interesting question! In computer vision, precision or accuracy is often used – and I also prefer to work with this metric. This and other metrics serve as a measure of the performance of a neural network model or algorithm, i.e. the percentage of correct predictions.
As a data scientist, what is your favourite tool to use?
PyTorch is definitely my favourite tool. It is an open source machine learning framework. We use this framework a lot in our work. PyTorch is deeply integrated with Python, so many Python debugging tools can be easily used with it. Also, like the Python language, PyTorch has an intuitive syntax and is relatively easy to learn. For those unfamiliar with the field, this may now sound very Spanish. 😊
With all the experience you gained at university, why say yes to Lindera every day instead of continuing research in academia?
I decided for myself that I think it makes more sense to apply technology to real products and add direct value to our society than to continue researching in academia. What we do today will not only help senior citizens, but also ourselves in the future.
Mobility means to you…
… the future.
What skills and qualities do applicants who want to join the Data Science team need to have?
We are looking for people who are creative and courageous and have experience in computer vision, e.g. with 2D/3D estimators and in gait analysis. If you want to work for us, you should have a desire to work in a team and interdisciplinary with different departments. At Lindera, we all see ourselves as a force working in the same direction. We welcome unsolicited applications! Simply send an e-mail to firstname.lastname@example.org.
Learn more about the Lindera team: