Python, C++, Java, TensorFlow, PyTorch Lightning, Automatic Speech Recognition, Natural Language Processing, CUDA, GCP, Facebook Flashlight, OpenAI Ray, OpenAI Whisper, Microsoft Deepspeed, Nvidia TensorRT, RAG, Google Vertex AI, Google TPU, Google Gemini, JAX, CLA, TPU

Python, C++, Java, TensorFlow, OpenCL, PostgreSQL, CUDA, PyTorch, Amazon AWS, Mongo DB, Kafka, Hadoop, NLP, ASR, STT, Anomaly detection

Implemented a real-time scoring system for heterogeneous source data; objects recognition system; voice chatbot using NLP; augmented reality prototype; banking mobile application. Modified 360° video processing browser plugin.

Used technologies: Python, C++, Java, C#, Javascript, TensorFlow, CUDA, Redis, PostgreSQL, Google Dialog Flow, Web Speech API, Android Studio, Swift, Natural Language Processing

Technologies: C#; C++; Accord.NET Framework; R; Python; TensorFlow.

Responsibility: Implementation of fraud detection system using AI.

1. Software implementation for nanochemistry reactions modeling. The project uses technologies of Distributed and Soft Computing, Machine Learning. It enables to create new algorithms for high-performance mathematical models calculation. Supported by grant from Russian Foundation for Basic Research (RFBR) 16-48-230433 in 2016

2. Implementation of innovative ideas of indexing and intelligent search based on methods of artificially intelligence and distributed computing. Supported by a grant from Russian Foundation for Basic Research (RFBR) «13-01-00807 А» in 2013

3. Forecasting4SMB project automates the analytical analysis of business information. It consists of two parts – subsystem for distributed high load calculations on idle office computers and cloud services, and subsystem for implementation of friendly and highly functional user interface. Supported by "IT breakthrough" http://tvoystart.ru/news/news-41717/ in 2011

4. Software implementation of model by combining ARIMA and layered full-connected neural network. Introduces a new way of coding categorical parameters. Introduces a new way of partially supervised learning of a fully layered neural network. Supported by grant from Russian Foundation for Basic Research (RFBR) "r2003ug" 03-01-96649 in 2003

5. Tesoro is an Open Source operating system project fully written in x86 Assembler language. The architecture is partially based on some QNX ideas and some fresh ideas. http://sourceforge.net/projects/tesoro/

6. Framework for search on large file systems. Cert #2013661245

7. Visual neural network function approximation. Cert #2013661246

8. Files search using neuro-evolutionary modeling. Cert #2013660776

9. Distributed statistics system for avalanches prediction Avalanche2003. Cert #2003610450