The evolution of Data Science at OLX: Insights from Tomek Jamiński

At OLX, data science is not just a department; it’s a driving force behind innovation and customer satisfaction. Tomek Jamiński, Director for Data Science at OLX, offers a fascinating glimpse into his role and the transformative changes within the company.

Can you provide a brief overview of your role and responsibilities at OLX?

Hi, I’m Tomek Jamiński, Director for Data Science at OLX. I consider myself to be the chief coffee maker in the company ;-) I’m a servant leader who sets the stage for all teams involved in solving customer problems using data. Specifically, I focus on creating an environment with the right balance of drive for results, support, challenge, and the human factor.

In practical terms, this means a lot of research on where to go next in terms of problem-solving, which technologies to onboard into our tech stack, and how to shape the development of our teams so that everyone can enjoy the ride. 

Taking care of people is essential for me as I truly believe that everyone should be the best version of themselves.

How has OLX's Data Science team evolved over the past few years, particularly with the integration of new technologies like Generative AI?

The world has become a lot more confusing as compared to what it used to be before the GenAI hype. Working with cutting-edge technologies means you are in a permanent beta test phase. Every week, something new pops up, and it’s very important to have a clear view of what is valuable and what is just shiny. We need to determine what can be used for internal productivity improvements and what can be used to solve customer problems by building impactful user-facing solutions. I think being able to play well in these two playgrounds is our secret sauce.

In what ways do you foresee Generative AI shaping the future of online marketplaces like OLX?

Putting the hype aside, I see many opportunities where GenAI can improve the way online marketplaces serve their customers. Problems that were difficult or very costly to solve can now be addressed more elegantly and efficiently. Take positing flow automation as one example: understanding images has become a lot easier. 

By leveraging GenAI models we can shift focus more towards the core of our business without bothering too much about particular building blocks. In the pre-GenAI era, we would need to invest enormous amounts of time into training advanced image models before being even able to demo a user-facing solution to stakeholders. That’s why having GenAI tools around is a big plus. 

From the user perspective, I can see that GenAI will help us reduce friction in our key user activities as well as information overflow. We need to simplify the trading process and make sure that we deliver more value than we initially receive from the user. The ultimate goal is to capture the user intent in the most intuitive and convenient way, and deliver results that are relevant and boiled down to the essence. Getting it right in the smallest number of steps possible is the challenge that we all face. 

What prompted OLX to expand its operations in Spain, and what are your strategic goals for this market?

We are a big player in the circular economy, which means a strong commitment to the environment and the community. We were looking for a place where we could create a Gen AI Hub and we know the talent and potential in Spain can really help us deliver this. We have been successful in growing talents across the markets we operate in, and we hope to repeat this success in Spain. The contribution to the local Data Science community is important for us as we love to establish mutually beneficial relationships.

With the introduction of new remote roles, how has OLX adapted its engineering workflows and collaboration practices?

Working remotely often means working in isolation. While this can be beneficial for maintaining focus, it also makes it difficult to build informal, personal relationships. Taking care of processes is easy, but we tend to put more emphasis on culture and human connections. We make sure that we know each other, not just work together.

Because the human factor matters so much we make sure that everyone gets the opportunity to meet the whole team in person at least once per quarter. This way we create an inclusive environment for those who mainly work at home, commute to the office, or operate completely remotely. 

What are your long-term goals for OLX’s Data Science team in the next 3-5 years?

Our long-term goals include adopting a “Silicon Valley” inspired, open way of thinking and fostering more structured innovation. This means running fewer initiatives but with a larger impact. We have embraced GenAI before it was cool and over the next years we will focus on establishing a mature environment that allows us to run GenAI at scale in a safe, observable, and testable way. We also remember about the “classics” as not every problem requires an LLM ;-) Sometimes a linear regression is enough to get the job done. Staying pragmatic is key.  

How do you foster a culture of innovation and continuous learning within your Data Science team?

We believe that initiatives should die on production and not on slides. The environment we are shaping supports innovation fast-tracks. We have codified it as the "3x3" approach, enumerating the aspects that are a “must have” and “nice to have” for every stage of the data product life cycle. In practice, it means that we expect people to be extremely fast in proving the point when doing a proof of concept, even at the expense of code quality. We want to ensure that bad ideas die fast so we invest more time into proper engineering standards once we know that things make sense from the customer's perspective. This way we avoid unnecessary work and we allow everyone to bring a great idea to life. The OLX Data environment supports the idea incubation process.