Fostering AI architects: a data strategy intern tells her story
As a Data Strategy Intern at idalab, Tina Emambakhsh explores how AI-/ML-based healthcare technology can be regulated in the future. Before joining idalab, Tina gained experience in a wide range of disciplines and sectors, having worked at the Austrian Embassy in Tehran, KPMG and Strategy&. Tina studied International Business Administration at WU Vienna and at Universidad del Pacífico, Peru, and is currently pursuing a master’s degree in International Economic Policy and Economics at SciencesPo and the Stockholm School of Economics. Her research interests lie at the interface between public policy and media.
How did you become interested in data science?
My first direct contact with data science started during a class about research methods, where I stumbled upon natural language processing and its use in analyzing the content of the World Bank’s annual reports on a lifespan of 70 years. As a language-enthusiast, I was fascinated by how analytically one can look at language given the vast amount of textual data available in digital form. Having an economic background, I immediately wanted to find a way on how to use the method for analyzing economic questions. So I started my own project where I collected articles mentioning the ECB before and after the Euro crisis in 2012 to analyze the effect of Mario Draghi’s signaling to the market on the composition and sentiment of European newspaper outlets. For this project I – as someone who had never seen a string of code before – started to learn programming in R, and really enjoyed it.
The results were astonishing: by only looking at the composition of words and their relationship, I clearly saw the narrative change in the media after he famously claimed to rescue the European economy by ‘whatever it takes’. However, what I also learned from my research was that learning from data can be heavily biased and results need to be interpreted with caution.
So it was really the exciting discovery from my research that set me off on the path to becoming a data scientist.
Interesting! And what do you find exciting about working at idalab?
While data strategy is more concerned with communicating intricate concepts to clients, I work closely with client company data scientists. I think that this kind of symbiosis is very unique and valuable given that it is not very common for consultancies to offer clients both high-level strategic guidance as well as tailored solutions to specific problems.
Data science as a discipline can be very intimidating at the beginning. This is further accentuated by the fact that it is abused as a buzzword in both media and business. But at idalab, you learn that data science is not just some sort of ‘magic sauce’ that you apply with the help of fancy data, but that it entails clear methodology and has certain barriers. In our projects, data strategists and data scientists carefully shed their respective lights on a client’s problem and analyse every necessary detail in order to fully understand the situation and subsequently identify the best solution. This mindset is also delivered to the client, which is one of the reasons why clients build a very close and trustful relationship with us, in my opinion.
What makes the internship at idalab so special?
It is definitely its hands-on nature and diversity of responsibilities. My main project is concerned with AI-regulation. During the very first week of my internship, I joined my colleagues in a client meeting where we discussed the implications of a recent FDA proposal on the regulation of AI-based medical devices. Now it is my responsibility to identify how such regulations can be implemented, among others, via the development of good machine learning practices. This is certainly a challenge, since a ‘gold standard’ for conducting data science has yet to exist. At the same time, entering these uncharted waters is also very exciting since this is a huge opportunity to generate tremendous value for the client.
Furthermore, I am engaged in several data science projects where I can dive deep into technical aspects of machine learning. In one of the projects, for instance, I support my colleagues in the conceptualization of computer vision-based medical devices. In another project, I have the possibility to try out data analysis myself. I especially enjoy the steep learning curve when discussing the use, practicability and limits of machine learning concepts with my data science colleagues. These conversations are worth more than any scientific paper that you could read and help to critically reflect on media buzzwords like like “artificial intelligence” and “digitalization”.
Can you elaborate on the topic you’re researching for idalab and why it’s so important?
My topic is mainly concerned with the regulation of artificial intelligence in medical devices. Different to traditional rule-based software, such as expert systems, machine learning technologies are heavily data driven and ‘learn’ from changes in their environment that is captured by the data they are processing. Hence, in the future, an ‘intelligent’ medical device will be able to change its behavior and tailor its services to the patient’s needs.
While this might sound useful at first glance, there are important risks to take into consideration before such devices can be approved for the market. The medical and public health industries are especially sensitive to risks posed by artificial intelligence since the lives of patients are at risk. Hence, current regulators, such as the FDA in the United States or the European Commission in the EU, aim to control the fast-paced nature of such learning devices by regulating their access to markets and defining good machine learning practices. In the far future, these good practices shall also be implemented in other industries that incorporate data science.
To sum up, what I’m doing now is conducting research on what kind of machine learning practices are already available. A very good source therefore can be found right here: namely our own data scientists. Given the vast amount of experience they’ve acquired in the field with different clients, it is very helpful to discuss scenarios of best practices that ensure safe and transparent design and development of machine learning algorithms.
Would you like to work as a Data Strategist in the future?
Data science is a field that continuously reinvents itself. As a consequence, Data Strategists will also need to keep growing and developing together with the changes of the discipline. For me it has always been key to work in an environment that enables personal growth and learning. And so I can definitely see myself as a Data Strategist in the future. I genuinely believe that the time at idalab has helped me to be more courageous with challenging my personal barriers- a skill that will be in growing demand in our generation’s nonlinear career paths.