Artem Mechetnyi

Research Engineer (Master's Thesis Project)

National Research Nuclear University MEPhI

A Russian national research university founded in 1942, known for nuclear science, engineering, physics, and technical education.

Tests deepfake model robustness against adversarial attacks, compares training defenses, and models attack success rates.

• Evaluating the robustness of deep learning models (Xception, MesoNet) by implementing and analyzing adversarial attacks (FGSM, PGD, C&W) on the DeepfakeBench codebase (PyTorch). • Implementing and comparing three training-time defenses (Vanilla AT, TRADES, AFSL) to study their effect on model robustness. • Debugging complex training pipelines, including normalization mismatches and loss collapse, across multiple model-defense combinations.

Research Engineer

Dagestan Federal Research Center, Russian Academy of Sciences (DFIC RAS)

A research center of the Russian Academy of Sciences in Dagestan, conducting work across natural and technical sciences through specialized institutes.

Explored ML for departmental research and developed a C# numerical solver for differential equations with singular coefficients.

• Studied machine learning methods and explored their applicability to the department's research tasks. • Developed a C# numerical implementation for solving a class of second-order differential equations with an unbounded (singular) coefficient, based on rational spline interpolation techniques. • Prepared the department's research activity report.

M.Sc. in progress, Information Security

National Research Nuclear University MEPhI

B.Sc., Mathematics and Computer Science

Dagestan State University