Student spotlight: Daniil Boiko

Lauren Smith

May 10, 2023

Three men look at one laptop in their research office

Source: College of Engineering

Gabe Gomes (center) and Daniil Boiko (right) in their lab

Ph.D. student Daniil Boiko's research goal is to enable chemists to do more complex things with fewer resources by using the power of data. With his advisor Gabe Gomes, Boiko is researching a diverse set of applications of machine learning in chemistry.

His primary project uses reinforcement learning for the optimization of reactional procedures. When chemists synthesize compounds, they must consider the order and timing of the steps in the process. Chemists also review the side products of the reaction and choose between catalysts. "This is a perfect problem for reinforcement learning," says Boiko. "It's all about an agent operating in an environment, doing some actions, and getting feedback from this environment." Ultimately, this approach could better optimize the synthetic route to any compound and reduce wasted resources.

Boiko is also working on new ways to represent molecules with graphs in machine learning algorithms. Accounting for the quantum nature of molecules greatly improves the performance of the model. "You can make these representations into models for toxicity, for drug-likeness, for many other downstream applications," says Boiko.

In another project, Boiko is developing algorithms that use data to engineer enzymes that can better catalyze reactions or perform new reactions. The project is a collaboration with a biocatalysis research group from the University of Michigan. Instead of training really big models, Boiko's approach relies on pre-trained representations of large enzyme sequences. Essentially, he is developing a recommender system for enzymes and compounds. A good frame of reference is the Netflix algorithms, which make recommendations based on assumptions about user behavior. The difference is that Boiko's model learns from experimental data, instead of from user interactions.

Growing up in Ivanovo, Russia, Boiko started studying chemistry as a path to medical school. In eighth grade, he was inspired to be a doctor when some family members faced significant health issues (which fortunately are now resolved). "I thought that if I want to be a doctor, I want to be a doctor in the best possible place in the country," he says. Admission to the medical school at Moscow State University is based solely on test scores. That year, it was particularly hard to get in, with a required score of 495 out of 500 across five exams. Boiko turned to the national school Olympics to bypass the normal test-based admission process. He became a subject winner in chemistry and then could be admitted to any Russian university without exams.


By his high school graduation, however, he realized that he was following a path because of events that happened in eighth grade and that "chemistry is cool." Instead of medical school, he enrolled in the chemistry department at Moscow State University.

Apart from studying chemistry and conducting research, Boiko also studied machine learning and data science in a center sponsored by VK. Some of the largest companies in Russia sponsor learning centers at the country's top universities, to strengthen their workforce recruitment pipelines. Boiko was invited for an internship at VK and then to stay on as a machine learning engineer in the search department. Among its services, VK is the second-largest Russian search engine. The machine learning engineers worked to create models that were as small as possible yet could perform as well as possible. "I got very good experience in software development, deploying machine learning models, and running them in production," says Boiko. "I could not have gotten this experience anywhere other than a company as big as VK." He particularly enjoyed seeing the direct impact of his work: that users had a better search experience from the models he rolled out.

As he neared graduation from university, Boiko wanted to continue his machine learning work on a larger scale and applied to chemistry. "Machine learning in chemistry is usually developed in chemical engineering departments," he says. The hardest part of getting to CMU was getting a visa, because of the war. The U.S. Embassy in Moscow is not issuing visas, and there are no flights between Russia and Europe. Last June, Boiko first flew to Dubai, then on to Belgrade to apply for a visa. As a technology expert, he was subject to additional security screenings. He waited for two and a half months, not knowing when he would hear the outcome. Finally, in late August, Boiko was issued a visa, and he traveled again through Serbia and on to Pittsburgh. Being so far from home is a continuing challenge. His mother calls every day to ask how he's doing.

Thankfully, the sense of community within the chemical engineering department has made a difference. "The work being done across the department is very different," he says. "For example, Anne Robinson and John Kitchin do absolutely different things. At the same time, there is the feeling that the department is one big, unified entity."

For media inquiries, please contact Lauren Smith at