● Full stack computer vision models development: - Labeling data using crowdsourcing platforms. - Model training pipelines using Pytorch/Pytorch-Lightning frameworks. - Serving ONNX/Torchscript models using NVIDIA Triton Inference Server. - Production-ready services development with Python/Golang. ● Developed from scratch, tested out and released models: - Model for checking the presence of a mask on a person’s face, inference with own C++ library on ATMs using the OpenVino framework. - Model for checking the presence of glasses on a person’s face. - Bestshot model – on-premise model for evaluating the quality of a photo (illumination, sharpness), inference with own C++ library on Android devices using Onnxruntime framework. - Model for recognizing a person’s gender and age. ● Developed from scratch an on-premise service for face recognition based on pretrained RetinaFace and ArcFace models. Projects: 1. CTF competitions platform. ● This platform was initially used for conducting information security classes in an IT summer camp for students. ● The platform is capable of handling multiple Capture The Flag contests in parallel. ● Implemented in Python language, using Django Rest Framework, Django Channels, Vue.js, MySQL, Redis, Nginx, Docker. 2. Implementation of FastGAN model. ● Implementation of FastGAN model using Pytorch/Pytorch-Lightning frameworks from “Towards Faster and Stabilized GAN Training for High Fidelity Few-shot Image Synthesis“. ● Notably, the model converges from scratch with just a few hours of training on a single RTX-2080 GPU, and has a consistent performance, even with less than 100 training samples. 3. Implementation of DeepSpeech2 model. ● Implementation of DeepSpeech2 speech recognition model using Pytorch/Pytorch-Lightning frameworks from “Deep Speech 2: End-to-End Speech Recognition in English and Mandarin“. ● Model is trained on LibriSpeech and LJSpeech datasets, LM-fusion with 4 gram KenLM model used in beam search ctc-decoding. 4. Implementation of FastSpeech model. ● Implementation of FastSpeech text to speech model using Pytorch/Pytorch Lightning frameworks from “FastSpeech: Fast, Robust and Controllable Text to Speech“. ● Model speeds up mel-spectrogram generation by 270x and the end-to-end speech synthesis by 38x compared to autoregressive models. 5. Implementation of HiFi-GAN model. ● Implementation of HiFi-GAN neural vocoder model using Pytorch/Pytorch-Lightning frameworks from “HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis“. ● HiFi-GAN generates samples 13.4 times faster than real-time on CPU and 167.9 times faster than real-time on a single V100 GPU with comparable quality to an autoregressive counterpart (WaveNet). 6. Implementation of SN-PatchGAN model. ● Implementation of SN-PatchGAN image inpainting model using Pytorch/Pytorch-Lightning frameworks from “Free-Form Image Inpainting with Gated Convolution“. ● Inpainting system is capable of completing images with free-form mask and guidance.

● Developed SOC-Robot – component of custom Yandex SIEM system. Robot is responsible for creating alert tickets in internal task tracking system, maintaining ticket lifecycle and launching additional checks. Robot is written using Python, Celery, Redis. ● Developed a system for static and dynamic testing of alerts with Docker and Splunk. Static tests allow triggering an alert on synthetic data and verify if it emits correct output. Dynamic tests allow end-to-end alert pipeline testing, launching an attack on company infrastructure and waiting until alert ticket is created in task tracking system. ● Participated in the alert development process with Splunk. ● Took part in second line security issues investigation.

● Analyzed the problem of uniform user data redistribution across multiple machines in scenario of distributed database cluster expansion. ● Came up with three different approaches depending on the cluster structure: the heuristic algorithm with O(n2) complexity for average size clusters, the probabilistic algorithm (uses SA technique) which works fine with large clusters and the precise O(n3) algorithm (Hungarian algorithm) which is perfect for small clusters. ● Implemented three algorithms of user data redistribution in C++. ● Developed administration and monitoring tools for data redistribution process.

● “Introduction to C++ programming language“ course teacher, two groups of students. Topics: variables, conditions and loops, functions, structures, std containers. ● “Introduction to information security“ course teacher, one group of students. Topics: web vulnerabilities, cryptography, steganography, networking, reverse engineering.