Interview with Petr Gusev, ML Tech Lead at Deliveroo and ML/AI expert

A exclusive interview with Petr Gusev, ML Tech Lead at Deliveroo and ML/AI expert

Petr Gusev is an ML expert with over 6 years of hands-on experience in ML engineering and product management. As an ML Tech Lead at Deliveroo, Gusev developed a proprietary internal experimentation product from scratch as the sole owner.

As part of the innovative stream of Yandex Music transforming the product to add podcast listening experience to the service, he built a podcast recommendation system from scratch as an ML Engineer at Yandex and achieved a remarkable 15% target metrics improvement. Additionally, as Head of Recommendations at SberMarket, his tech-driven roadmap elevated AOV by 2% and GMV by 1%

Can you share a bit about your journey in the field of machine learning and artificial intelligence, leading up to your current role as the ML Tech Lead at Deliveroo?

Petr Gusev: Prior to Deliveroo, I had already accumulated substantial experience in machine learning and artificial intelligence. Part of this experience came from the consulting world, where I was an ML Team Lead, tasked with solving problems by applying ML to different domains like real estate and e-commerce. 

I also worked for two leading companies. First, I was a machine learning engineer at Yandex Media Services, where I built a recommendation platform that adapted to user preferences in real time, and provided them with music and movies streaming recommendations. Then, I became the Head of Recommendations at SberMarket, where I led the transformation of the company’s consumer app by placing a strong focus on personalisation. 

As someone with a background in both engineering and product management, how do you navigate the intersection between these two domains in the context of machine learning projects?

Petr Gusev: It really helps me to be able to balance the execution speed while managing stakeholders’ expectations. My product manager background helps me understand the problem we are trying to solve for the business, and what business (not ML) metrics are we targeting by implementing this ML approach. After I understand the entire context, I come up with an ML solution for the problem, experiment and see if our business goals are feasible with the solution I’ve come up with.

Sometimes, I come up with some interesting solutions that don’t really require complex ML models, so we need to be able to iterate very fast, which is always a fun part 🙂 

Deliveroo operates in a fast-paced and dynamic industry. How do you approach the challenges of applying machine learning to enhance the user experience and efficiency within the food delivery ecosystem?

Petr Gusev: My work at Deliveroo is under an NDA, so I will share my experience from SberMarket. In my opinion, in a company that actually operates in the offline world, (i.e. it has riders who are travelling, customers who are buying physical goods, etc.) one of the biggest issues is data quality. To address this at SberMarket, we devoted a significant amount of our time to building and maintaining an ML data lake, which contained data both from the back-end data and events.

We documented all the tables we had in the data lake and added monitoring dashboards, so everyone using the data would know where it came from and how it was collected. This allowed us to faster prototype new ML models and measure the potential impact of the features. 

The second thing we did was to integrate and implement ML Ops tools, which helped us to build and ship new models faster. 

Last, but not least, we emphasized clear task formulation and a continuous process of gathering insights from our users. We had a group dedicated to customer research, and those who were a part of it would constantly perform user interviews and gather insights. Also, we found it extremely useful for at least 1 MLE to shadow our researchers. The amount of insights they would always come up with is enormous, and we were able to make several successful implementations based on those insights. 

As a tech lead, how do you foster a collaborative and innovative environment within your team, especially when working on complex machine learning projects?

Petr Gusev: Especially when working on complex machine learning projects, collaboration is key. In order to foster this collaborative environment, everyone needs to feel a part of the team. Therefore, I ensured that everyone felt seen and heard, and knew that their opinion mattered. Also, I developed strategies that helped creativity to flourish.

For example, every once in a while, I would rotate MLEs around the different areas of the project. This expanded both knowledge and awareness for every team member, and resulted in new, innovative ideas. 

In addition to this, I paired junior engineers with senior MLEs and assigned them to specific research projects. This helped us to bring junior MLEs up to speed and to learn from those who were already more experienced. We also organized offline hackathons, which were great for the team to try new things in an environment where they felt safe to think out of the box. 

Can you share an example of a particularly challenging machine learning problem you and your team have successfully tackled at Deliveroo, and the lessons learned from that experience?

Petr Gusev: My work at Deliveroo is under an NDA, so I will share my experience from SberMarket. Once, we were implementing a technology called AI-driven UI, which consisted in us ranking and placing various elements of the user interface in a personalised manner in order to design a better experience for our users.

There were many lessons learned from this. To begin with, we learned that before developing any UI feature, we must find a way to gather data in regards to how these modifications affect users’ interactions with our app. This way, we can get feedback as to what worked and what didn’t.

Also, it is very important to start with the simplest solution possible, and to gradually increase the complexity of ML models based on results and findings. To do this, it is key to have a cross-functional team that includes both client-facing individuals and back-end developers that are familiar with ML-heavy components. 

How do you stay updated on the latest advancements in machine learning, and how do you encourage continuous learning within your team?

Petr Gusev: I love to be engaged with the broader ML community and frequently attend local events, for example, Data Breakfasts, where many ML practitioners gather to discuss ML projects and innovations. I encourage my team to do the same, and also to attend conferences and meetups, so that they can expand their network and stay on top of what is going on.  

What advice would you offer to aspiring machine learning professionals looking to make a meaningful impact in both engineering and product management roles?

Petr Gusev: I believe that in order to solve problems in the best way possible we have to step into both roles, and look at the problems we are solving both from the MLE and Product Manager point of view. So everytime I solve a problem, I ask myself two questions: 

• What problem am I trying to solve for users? 
• How should I solve it in the most efficient way?

I truly believe that the first question should come first, because the answer to the second question really depends on how we answer that initial question. To progress in both fields and be successful I think one of the most important things is observation. My advice would be to follow some inspirational leaders from both Product and Engineering, read their blogs / newsletters and attend conferences.

It boosts your observation capacities in a very healthy way, because now, whenever you try to solve a problem, you already know how the industry leaders solved similar situations, and you will implement the best solution possible. This boosts your creativity and innovation.   

What do you see as the future of machine learning and AI in the food delivery industry, and how do you envision it shaping the overall customer experience?

Petr Gusev: I think the way in which AI will transform the food delivery industry is by enabling the hyper-personalisation of the user experience. We all are really different when it comes to food and we all have different purchase patterns. For example, some people have dietary requirements, some people like to purchase items in bulk for the entire week, some of us just occasionally pop up in the store and buy what we want just right now.

Or, sometimes, the same person can have different behavioural patterns depending on the time of the day or day of the week.  Currently, not many apps detect and support all of these experiences. With the help of AI, we will be able to learn these patterns from our users and support them in order to help them achieve their goals in a very convenient way.   

What is your success tips for young and aspiring entrepreneurs

Petr Gusev: I would encourage entrepreneurs to test as many hypotheses as possible. I believe doing this will play a key role in your success. Because, being realistic, you cannot develop a product from scratch faster than a big corporation that has more capital and resources than you. However, you can be faster when pivoting and snatching opportunities, so having this agility will help you do that. 

Then, do not overcomplicate things. Simplicity far outpaces complexity. I also recommend that you stay on top of what is going on in the industry. By being up to date with the latest technological advancements, you will be able to adapt your product to what is happening right now. 

Last, but not least, just try and break things 🙂 There are many ML tools available right now, so don’t be afraid to try them out, see how they work, and experiment with them. By recording your observations, you will be able to identify how you can apply one of these ML tools into your business, regardless of whether it operates online or offline. 

Read more Success stories of Indian entrepreneurs, Women Entrepreneurs & startups stories at SugerMint. Follow us on Twitter, Instagram, Facebook, LinkedIn