The Botkin.AI project is named after Sergey Botkin, one of the founders of modern clinical medicine.

In his philosophy, Sergey Botkin viewed the human organism as a single whole that has an indissoluble connection with its environment. Botkin’s clinical concept is built upon the teachings on the internal workings of how pathological processes develop in the body.

Botkin’s dream to devote his life to mathematics is a well-known fact in his biography; only circumstances prevented his enrollment in the Faculty of Mathematics at the Moscow University. Botkin.AI, the name of our mathematical patient modeling technology, is a tribute to the great doctor, who had a dream of becoming a mathematician.


Our products utilize a unique technology to build mathematical models for patient representationKey elements of our technology:


Botkin.AI performs domain-specific data preprocessing that depends on:

  • Data types (images, 3D images, video, 3D video, signals (time sequences), tables, text)
  • Data sources (EHR, X-ray, CT, MRI, ultrasound, mammography, ECG, EEG, blood/urine/stool/CSF tests, histology).

Even without manual labeling, the platform automatically prepares a machine learning dataset from different types of data:

  • Medical images;
  • Genomic data;
  • Medical test results.

Botkin.AI automatically retrieves facts from the patient’s health records, highlights diagnoses and other relevant facts, places them on a timeline, sorts them by examination and forms {multitude of facts – diagnosis} pairs from the groups


The platform automatically trains models on available selections of preset parametric model families for every type of data. Parametric families determine the general look of models with a certain selection of parameters, which are strictly defined by the model


Final classifiers are trained together with the aggregating model for the purpose of

  • diagnostics
  • disease
  • progression analysis
  • treatment recommendation

The platform uses the aggregating model to build a vector model of the patient from all the available data, evaluating the relevance of every fact for the given diagnosis or group of diagnoses


Medical ontologies enable Botkin.AI to:

  • Map raw text onto medical terms at the preprocessing stage;
  • Limit the structure of the model’s environment during the learning process;
  • Regulate joint representation training