HR Talks with IT Leaders ep. 8: Iva Gumnishka on building an impact-driven team, gender inclusion, and removing bias from AI
HR Talks with IT leaders is a campaign organized in collaboration between BICA Services and the Recursive, one of the most prominent media providers in the Bulgarian market. Our goal is to give more visibility to the knowledge of how great tech teams are built. We meet with accomplished entrepreneurs and managers who share their personal experience and what’s their approach to leadership, communication, hiring, talent development, and much more.
In our next conversation, we talk to Iva Gumnishka, founder and CEO of Humans in the Loop (HITL). This is a social impact startup that helps refugees, asylum-seekers, and migrants gain in-demand IT skills such as the collection, annotation, and validation of the data needed for bias-free AI models. In 2020, HITL went on to become the first Eastern European company to win at the MIT Solve challenge, getting two awards, including the Andan Prize for Innovation in Refugee Inclusion by The Andan Foundation and the Gulbenkian Award for Adult Literacy by the Gulbenkian Foundation. HITL has also supported numerous prominent AI solution builders such as Imagga, Daedalean, Edamam, Alcatraz AI, and Sensika.
What’s the number one lesson about leadership you learned over the years with Humans in the Loop?
I’ve always tended to consult a lot with people when making a decision, you know, looking to collect different types of inputs from everyone willing to offer. At one point I considered changing this and starting to make decisions more autonomously. I shared this with the team and they all told me not to change it because it’s one of the best things about me. Rather, it’s about collecting everyone’s feedback, choosing one direction and sticking with it, even if some people aren’t happy about it.
What skills and mindsets are you looking for when hiring?
For all the people in the company, we’re looking for a commitment to our social mission. We want people who believe in what we do and who can adopt our model and embody it in all of their work. It’s really important that all of our new hires understand our social impact and represent it in all of their business decisions.
We recently hired a Chief Commercial Officer who’s now leading all of our sales. He’s been great in transmitting our social impact to potential customers and he’s always ready to say no to a client who tries to negotiate towards a price that will push us below what we’ve defined as decent salaries for our workers. From a traditional sales perspective, he might have been frustrated that he’s unable to close as many deals because of our minimum payment rates, but he is defending our values in all of his daily work.
When it comes to our trainers and people working with our beneficiaries in general, I’d say that most of all, they should be able to communicate in a very clear language. Lots of the people they will be working with are coming without prior experience or domain knowledge. So it’s really important for the trainers to be able to transmit their skills, to recognize the talent in each person, and find a way to bring it out.
Tell us about a hiring mistake you have made and what you learned from it?
In the beginning, I had no hiring experience. I hadn’t even interviewed for a job as starting Humans in the Loop was my first job.
One of my very first mistakes was not reviewing the entire pool of applicants. You have a job advert open, people apply, and you start interviewing them but the job ad is still open and people continue to apply. It’s a really rookie mistake if you don’t organize your hiring process properly so as to give everyone an equal chance, no matter whether they applied in the first few days or a bit later.
That one was easy to fix but a more serious mistake was hiring consultants instead of full-time team members. Because for a startup, it’s really important to have people who are dedicated 100% to the company and ready to do all the grunt work that’s involved in getting a company up and running.
We used to have very good candidates applying for our core team but their rates would be super high — I could never afford to offer them what they’d be paid in another place. So we reached a consensus that we’d hire them as consultants who’d get paid for 8–10 hours per week.
This turned out to be very harmful to the organization in the long run as we ended up not having a team but just a bunch of consultants who’d give very thoughtful advice but there’d be nobody to execute it. So I would end up doing all the tasks coming from all directions.
We lost a lot of time — between six months and a year before I changed my mind and decided not to hire consultants ever again. It was far more important to hire people with the right values who were able to join full-time and build the organization together with me.
What soft skills do you teach to refugees in order for them to be effective freelancers and members of the workforce?
In order to get soft skills it’s not enough to just sit in a classroom, you have to actually experience real-life situations which will shape your professional behavior and habits. When we employ refugees directly, we provide them with the opportunity not only to make some additional income but also to learn how to work in a team, how to work remotely, how to manage their own time, how to deliver with good quality, how to be accountable for their work, how to take initiative, and how to communicate when they’re unsure what to do. There is no way to teach all of these soft skills in theory, you have to learn them in practice, and the organization’s role is to set the right expectations towards you and to provide you with the right guidance and mentorship.
One of the items in HITL’s fair work policy says: “We are committed to gender equality and strive to ensure that we have at least 50% women in our workforce”. Tell us more about gender inclusion practices and how do you live up to them?
On our core team, 7 out of our 10 people are actually women, including 3 out of 4 people on the management team. Gender inclusion is a main priority for us as an organization and it starts at the very top. But that’s not all of it: we have 7 different nationalities on the team with 3 continents represented, and we are also very diverse in terms of age. So, it’s not just about gender but making sure that we represent a wide variety of viewpoints and experiences, in all types of different dimensions.
When it comes to our annotators and beneficiaries, it’s actually quite easy to achieve gender balance and a high level of representation of women in most of our locations — Bulgaria, Lebanon, Afghanistan, Iraq, Yemen. This is because we are implementing a lot of strategies to make the type of work we offer attractive and suitable for women.
In general, remote work and work from home are very attractive for women as it allows them to manage their own time and to balance it with other responsibilities like family and childcare. We’re really trying to adapt our deadlines to make sure women have enough time to work, which is a challenge in some of our locations like in Syria. During in-person trainings, we’re also striving to provide childcare, so women can have someone to take care of their children.
What do you do to remove bias from AI models?
One of the main focus areas is dataset collection. When you’re collecting data you need to make sure you’re representing lots of different regions, a lot of different demographics, ethnicities, and so on. Every time when we have to collect a new dataset, we do a thorough analysis of what diversity dimensions will be necessary and how feasible and realistic it would be for us to acquire this data.
When it comes to annotation, bias is very often related to the diversity of opinions different people might have when labeling the data. Usually, the industry is centered around the consistency of the labels and any type of outliers are penalized, for example a labeler with a less standard interpretation of the data. However, we pay special attention to edge cases which represent this type of disagreement between labelers and discuss in detail with the team and clients how they should be interpreted.
Moving forward, how would you measure your success as a leader? If we talk again in 2024, what must have happened till then so you can honestly say to yourself that you’re happy with what you’ve achieved?
For us, 2024 it’s an important horizon as we’re trying to establish ourselves as the leaders in the market for real-time monitoring of AI systems. As an organization, we’ll be looking to tap into this business niche and prove that our impact model can scale — that people can come to us, receive training and work experience and then move on to other opportunities. We’re currently working on defining this employee journey and exploring how we can further support them in their future professional and personal development. If we prove this model, which is to be piloted in Bulgaria, and we can confidently say that we’re impacting lives for the better and achieving better outcomes for refugees, I think this will be a success for me.