National Research Nuclear University MEPhI (Moscow Engineering Physics Institute)

Software Engineer with Machine Learning experience.

HeadLight

Machine Learning Engineer

- Developed from scratch a machine learning model for DDoS-attacks clusterization using hierarchical clustering that provided new insights into the Security Operations Center. Gave a talk at the AI Journey, the most credible AI conference in Russia, presenting results of the project.
- Developed a confidential document classification model using TF-IDF features and gradient boosting decision trees that improved baseline F1-score from 0.71 to 0.93 and beat third-party model on a holdout data by 0.05 precision at (recall=0.95).
- Implemented a graph-based approach to analyze e-mail logs, which helped to reveal new channels of information leaks. The results were presented to the CISO of Sberbank.
- Developed a  machine learning model based on gradient boosting decision trees that predicted operational costs for buildings at the end of the year, it allows evaluating cost overruns.

Sberbank

- Developed a biometric testing framework.
- Wrote Python’s wrappers for C/C++ biometric APIs using Cython and pybind11.
- Conducted tests for vendor’s biometric solutions, evaluated FAR/FRR metrics.
- Integrated the prototype of the fingerprint authentication system (my internship project) into the production environment:
- provided integration with LDAP system.
- developed a web-interface using Django.
- set up automated builds from a Dockerfile.
- added fingerprint data encryption and integrity verification.

Software Engineer

- Analyzed existed open-source biometric solutions for fingerprint verification and selected the best one.
- Created a prototype of fingerprint authentication system based on NIST Biometric Image Software.
- Tested the prototype on open fingerprint databases like NIST, CASIA, ATVS-FFP and evaluated FAR/FRR metrics.