Machine learning (ML) has garnered increasing attention from businesses in recent years. As of 2023, drivers behind this trend include extracting higher-quality information (60%), increasing process productivity and speed (48%), reducing costs (46%), and extracting more value from data (31%). Gleb Sinev, a product owner at a Europe-based FinTech company, encountered versatile applications of ML in a broad range of industries in his career path and believes ML could change the business landscape much more significantly in the future.
Personalized Content Delivery
Gleb initiated his engagement with machine learning in his work as Project Manager starting in 2015 at Surfingbird, an advertising company that utilized ML to suggest personalized B2C content based on users' interests and browsing history, aiming to boost engagement.
ML was really helpful in solving the "cold start" problem in recommender systems by using demographic information from user registration, like gender, age, and location, to make initial recommendations for new users or sites.
As Gleb notes, by clustering users based on this data, the product provided accurate suggestions without needing extensive user history.
Another product he was responsible for as Project Manager at the company was Relap.io. It targeted the B2B sector and functioned as a native advertising platform. In 2016, it was integrated into the entertainment website Adme.ru, which boasted a monthly audience of 17 million visitors.
"After integrating our solution, we observed a 33% increase in readers clicking on links within the recommendation block. Moreover, the deployment of Relap.io widgets on the news agency site ria.ru resulted in a doubling of the click-through rate (CTR), which underscored the substantial influence of Relap.io's technology in captivating audiences and improving user interaction," he highlights.
Furthermore, Surfingbird obtained official whitelisting from Adblock Plus, a widely utilized web plugin, guaranteeing that the majority of our ads remained unblocked. This initiative Gleb conceived, developed, and implemented in collaboration with the programmatic advertising team. The outcomes were remarkable: certain clients witnessed a notable revenue surge.
"Jumping ahead, I departed from Surfingbird in 2018 but returned in 2020 at the founders' invitation. During this period, the company was acquired by the IT giant Mail.ru (now VK Holding). Upon my return, I was entrusted with the responsibility of integrating Surfingbird's advancements into Mail.ru's proprietary recommendation system, Pulse," recalls Gleb.
The main page of Mail.ru was a part of this platform, offering users personalized content, among other things, and generating millions of views a day. However, prior to Surfingbird's integration in 2020, the page operated under manual control. The primary objective of Mail.ru was to transition it into an automated content recommendation system, tailoring recommended articles and advertising materials.
In his role as a Lead Product Manager, Gleb led the work of three dynamic teams: the ML teams at Pulse and Relap.io, in conjunction with the infrastructure team at Pulse.
Gleb led the Surfingbird team through product testing before branching out the platform. This expansion led to the development of specialized tools designed specifically for editorial teams. They created innovative solutions that enabled editors to direct traffic toward selected publications without the need for IT expertise. Included in these advancements were features aimed at increasing the visibility of specific content to drive traffic. Over two years, Gleb's initiatives resulted in a significant increase in click-through rates (CTR) and the successful integration of personalization features.
From Document Digitization to Cattle Breeding
Another notable experience Gleb had was his tenure at the ML startup Dbrain (now Handl.ai), which was a participant in Y Combinator, a renowned startup accelerator. He joined the team as a Product and Project Manager.
Initially, the company began as a blockchain ML venture; however, soon, the startup recognized the need for a change in direction, finding a market fit in the field of document conversion, including invoices, income statements, and custom forms, leveraging computer vision technology. This involved a blend of cameras and AI-based software.
The company dedicated its resources to the development of DOCR, a SaaS solution focusing on optical character recognition (OCR) and data extraction, designed to automate any document-intensive business process. This included the sophisticated recognition of both machine-typed and handwritten documents.
This initiative aimed at harnessing advanced OCR technologies to streamline the process of converting various types of documents into editable and searchable data, thus facilitating efficient information retrieval and management for businesses across multiple sectors. Handl began offering it to financial companies and banks, where it was selling exceptionally well; at the same time, Gleb's team continued to improve and refine the product, looking for new market fits.
The proprietary data annotation platform was developed and equipped with assessors for image, sound, and text data recognition. Gleb was responsible for setting objectives and orchestrating teams' work. Additionally, he handled negotiations with investors and clients while focusing on the creation of bespoke solutions. Subsequently, the product gained traction among various companies, including Nestle, AMG, and Ligolab.
One of Handl's most thrilling pilot projects was undertaken for a significant agricultural holding. The development team employed cameras and ML solutions to meet their requirements. For instance, when a cow falls ill, its behavior shifts—it may move less, drink more, sleep more, eat less, and alter its interactions with other cows. So, the system was trained to differentiate between cows and other objects, such as haystacks, and to detect each individual animal, leveraging their unique coloring patterns.
Gleb spearheaded this development, focusing on a machine learning application that identified changes in cow behavior indicative of illness. This test project allowed Handl to test various hypotheses and refine algorithms, showcasing the use of Computer Vision in agriculture.
Another one-of-a-kind marketing product based on Handl was built for Dodo Pizza, an international pizza delivery and takeout chain that operates in Europe, the Middle East, etc. Machine vision was utilized to assess the doneness of pizzas based on their crust.
"The process was quite straightforward. As the pizza cooks, the crust changes color and develops distinctive, slightly darker bubbles. By analyzing the number of these darker spots on the crust, we can determine how well the pizza is cooked. For the application that recognized the quality of the test, more than 50,000 photos of pizza made from different foreshortening were marked," says Gleb.
ML in Healthcare
In 2021, Gleb Sinev was approached with an opportunity to apply his knowledge in creating a healthcare product with ML by the founders of Mantika.ai. They were searching for a market fit for their newly established company.
"In healthcare, ML and artificial intelligence play pivotal roles in tasks such as diagnosing diseases more accurately, predicting patient outcomes, and devising personalized treatment plans," notes Gleb. "Recognizing the global significance and devastating impact of lung cancer, we decided to focus on detecting it in CT scans. Given its widespread prevalence and high mortality rates, addressing this issue held immense potential for reducing both human suffering and economic burden. I was responsible for overseeing the entire project at the operational level, including assembling the development team. I conducted market research, analyzed the competitive landscape, and proposed several options for our focus."
Mantika.ai successfully launched a product featuring an intuitive interface. It enabled users to identify suspicious areas in CT scans and facilitate report downloads through a web interface. Shortly thereafter, partnership agreements with five major clinics to implement the solution were forged.
ML and Product Management: Forging Ahead
Machine Learning has transitioned from a mysterious and rarely used technology to something accessible and cost-effective for businesses. Even small businesses like cafes can implement simple ML solutions, such as Quality Control for food preparation, at a low cost.
Gleb is convinced that the increased accessibility means that Product Managers should have a solid understanding of ML as it becomes one of the new essential tools, and the knowledge base required for product management is expanding accordingly.
"The hype surrounding machine learning is sometimes substantial, and with it comes lofty expectations. It's important to understand that machine learning isn't a magical solution that can create results from thin air. Rather, it's a tool that, when used appropriately, can yield significant benefits. Still, just like any other tool, its true value is realized only when it's employed in the right context," he notes.
