Dr. Larysa Visengeriyeva, Head of Data + AI at INNOQ and founder of Women+ in Data and AI © Private

05 August 2024

“We want to change the disbalance on female leadership in data and AI on the systemic level.”

Women in the data and AI industry remain underrepresented, leading to lower female participation rates, fewer leadership roles, and less visibility in high-profile projects. The lack of representation not only limits diversity of thought and innovation but also perpetuates gender biases within AI systems themselves. Increasing the presence of women in data and AI is crucial for fostering a more inclusive industry, promoting diverse perspectives, and developing fairer, more equitable technologies. A leading figure in the fight to bridge the gap and ensure a more balanced representation in the future of tech is Dr. Larysa Visengeriyeva, who is the head of Data + AI at the technology consultancy INNOQ and founder of Women+ in Data and AI. Ahead of the festival’s second edition in Berlin, we talked to her about her beginnings in data and AI, women’s many challenges in the industry and how a festival can create a movement for positive change.

Hello Ms Visengeriyeva, thanks for taking the time to talk to us. Could you share your personal journey in the field of Data and AI?

Thank you for inviting me! I explored the field of data during my thesis while completing my graduation. It focused on benchmarking persistence frameworks such as Hibernate, which was a hot topic back in 2007. After starting out as a data engineer, I made a career shift into academia by joining the Database Systems and Information Management Department at TU Berlin. My AI journey began in 2011 when I became curious about machine learning after hearing about it at conferences.

Initially, I worked on NLP-based research involving machine learning techniques. However, my AI journey took a new direction when I switched my focus from NLP-based research to addressing data cleaning problems. This transition was further strengthened during my PhD on Augmented Data Quality, where I extensively used machine learning. Eventually, I developed a keen interest in the productization of AI, and by the time I transitioned into the industry in 2020, I was already aware of the exciting and important field of MLOps, which led to the creation of ml-ops.org.

Currently, my focus lies on the operational aspect of EU AI Act compliance.

What inspired you to establish Women+ in Data and AI? Can you share a specific moment that comes to mind? And why did you choose Berlin as your base?

On June 27th, 2022, in Berlin, I gave a talk at the Women+ in Machine Learning and Data Science meetup. The focus of the talk was demystifying MLOps and explaining its principles, importance for delivering business value, and navigating the field of MLOps tooling. The event had an inclusive atmosphere, bringing together various female tech communities, including WiMLDS, PyLadies, and Women in Robotics and AI, and featured a group of startups with MLOps expertise. The Q&A session and discussions covered topics like Machine Learning Governance and the environmental sustainability of resource-intensive ML models.

This experience was transformational for me. It was the first time I felt a sense of safety and respect while speaking in public. Everyone was eager to learn, and it was truly empowering. The meetup provided opportunities for networking and connecting with female tech communities, which further boosted my sense of empowerment.

This experience encouraged me to consider the idea of organizing a tech festival to celebrate female tech excellence. I wanted to scale the concept of WiMLDS by providing a platform for female+ speakers and tech communities specializing in data engineering, data science, MLOps, and AI in production to come together and celebrate. This is the empowerment I want for every woman+ in tech.

In your view, what are the most significant challenges that women face in the Data and AI industries today? And have you seen steps towards a better environment?

I believe some of the most significant challenges for women in Data and AI industries include the lack of visible role models: Women make up only around 26% of data and AI positions in the workforce, despite being 47% of the overall labor force. This lack of representation led to feelings of isolation and limited access to networking, mentorship, and role models for women in the field. With so few visible women, especially in leadership roles, it's harder for young women to envision themselves in these careers.

Looking at some data and AI workplaces, they can have a strong "bro culture" and overall toxic culture that isn't welcoming to women. This can be an isolating environment that discourages women from staying in the field. No wonder that many women in tech, data science, and AI struggle with self-doubt and imposter syndrome despite their accomplishments. We question our skills and whether we belong in the field.

Women often have more limited access to high-profile projects, leadership positions, and career advancement in data science and AI. The pay gap between men and women in the field also remains significant. These disparities slow women's progress and contribute to the underrepresentation.

Finally, women regularly experience prejudice that undermines our credibility and technical abilities. There is still a strong belief that fields like AI and data science are more suited for men, leading to skepticism about women's ideas and expertise. A study by the World Economic Forum found that women are 25% less likely to be hired for AI roles than men, and once hired, they are 15% less likely to be promoted. This bias is rooted in traditional cultural stereotypes that boys are better at math and STEM.

To address these issues, we need change at the individual, educational and industry levels. Women must advocate for ourselves, seek out mentors, and continuously expand their networks and the AI knowledge. Companies must adopt inclusive hiring practices, promote women to leadership roles, and actively work to improve diversity. Most importantly, women should proactively claim their places and roles at all levels.

How can the festival be a part of positive change?

At the W+DAI festival, we have observed a significant impact. One noticeable outcome is the collaboration among female tech communities that connected during the festival. In the last and this year, we have seen a number of joint meetups that have been organized by different communities after the festival. These intersections of communities are essential for expanding our network and promoting a sense of unity among women in tech.

In addition, we have had the privilege of supporting first-time speakers on their journey to becoming public speakers, and this has had a positive impact on their professional careers.

Furthermore, we noticed the mentorships originated from the inspiration provided by the speakers at the festival. The creation of a "safe space" at the festival has empowered attendees to ask for mentoring, which established valuable mentor-mentee relationships.

I am incredibly proud of the Leet Ladies Ljubljana, who attended W+DAI 2023, got inspired by the idea of the festival and created a similar conference in their home city. DATA_FAIR got me enormous motivation to continue on the W+DAI festival in 2024. You can read about their story here.

And I am pretty sure, there was much more happening that I am not aware of!

What are your plans for the 2024 edition of the festival and what can we expect from it?

This year's festival theme is "Women+. Entrepreneurship. AI" to encourage more women+ to build their own companies and therefore change the disbalance on female leadership in data and AI on the systemic level. We are teaming up with BHT Startup Hub to arrange a panel discussion on this subject. We are also working in partnership with the SIBB (Association of ICT & Digital Business Industry in Berlin and Brandenburg) Incubator, who are committed to supporting Women in Tech and early-stage tech startups. A workshop on "Disciplined Entrepreneurship" is also in the pipeline. Furthermore, we plan to include an activity titled "(Female) AI-Co-Founder Search" as part of the festival.

In addition to the W+DAI Festival on September 27th, throughout the year, we are extending the festival by hosting a series of satellite meetups in collaboration with various companies and communities, such as AICamp, SIBB, Thoughtworks, Red Hat. These meetups are called W+DAI Nights and are smaller versions of W+DAI. They follow the same concept—providing a platform for women+ to take the stage while welcoming everyone.

And as usual, you can expect a great environment full of tech discussion, grow and inspiration, and empowerment for everyone.

While gender diversity is the focus of the conference, it hasn't been discussed in any of last year`s talks. Can you elaborate on that choice?

We don't talk. We show, we create, and we live diversity!

We've established a unique space where women can showcase their expertise, foster idea exchange, connect with other experts, inspire and be inspired. We aim to make this the norm to see women leading technical discussions.

As a woman in data and AI, I don't need to be reminded about all diversity challenges. I want to increase my network, inspirational discussions, role models, and tools to achieve my goals. I have a personal experience when I got a diversity scholarship for writing my PhD Thesis and had to attend a seminar as a prerequisite. That seminar's topic was - "Diversity". Imagine a room of 30 female PhDs-to-be sitting there and listening to why diversity is important. That was a moment when I realized that instead of spending our time on what is absolutely obvious to us, I would rather spend my time learning how to invest, how to become a public speaker, create my personal brand, or something that would give me a tool to level up.

How does increased diversity, particularly gender diversity, enhance the development and ethical implementation of AI technologies?

Sure! Let's take a step back, sit down, and read "Invisible Women: Exposing Data Bias in a World Designed for Men" book, published in 2019 by British feminist author Caroline Criado Perez. This book reveals the ubiquitous gender data gap in our society and its far-reaching consequences for women. This leads to AI systems that incorporate gender discrimination and fail to adequately serve women's needs. Increasing gender diversity in AI teams helps identify and mitigate these data biases, leading to fairer, more inclusive AI products. Gender-diverse AI teams bring a range of viewpoints and experiences to the table. This diversity facilitates creativity and solutions considering a broader spectrum of societal needs.

Increased diversity, specifically gender diversity, is crucial for implementing AI systems. Inspiring the next generation is important, which is why role models and representation matter. Seeing successful women in AI encourages more girls and young women to pursue STEM education and careers. This diversity creates a positive flywheel effect - the more women shape the future of AI, the more inclusive and ethically grounded the field becomes.

Could you discuss the impact of gender bias in AI algorithms? What steps can be taken to mitigate these biases at the developmental stage?

Gender bias can be manifested in data, algorithms, and people (teams). As I already mentioned, several steps can be taken at the developmental stage to mitigate gender bias in AI algorithms. Remember to make sure that the training data for AI models includes a balanced representation across genders and other sensitive variables. It's important to collect more data from underrepresented groups and to consider how gender intersects with other attributes like race.

Having more women and underrepresented groups involved in designing AI systems is a good idea. This can help identify blind spots and bring in more inclusive perspectives. It's also recommended to have cross-functional teams with gender and ethics experts.

Before deploying AI models, it's essential to frequently audit and test them for gender bias. Measuring performance separately for each gender is necessary to identify any disparities. It's also important to have transparency around AI decisions.

Techniques like counterfactual fairness to reduce dataset bias should be applied to ensure inclusivity. Careful attention is needed in problem formulation and defining target variables to avoid reinforcing stereotypes.

Furthermore, according to the EU AI Act, non-discrimination should be guaranteed, and proper documentation of ML models should be provided.