Evgeniy Shashmin

- Developed GPU and NPU virtualization software (library-level, driver-level) for cloud and data storage solutions for AI applications. - Wrote C/C++ code generator in Python to produce customised HPC libraries (CUDA, CANN). - Reverse-engineered non-public parts of CUDA API. - Implemented modules for workload management and scheduling. - Researched new directions to improve hardware utilisation for diverse ML workloads, presented tech. insights at TED meetings, wrote SoW for academic coop. projects.

- Wrote C++ backend code for smart display (AI voice assistant product family) to support core features/scenarios (weather, traffic, video, music and others). - Cooperated with other teams to enable integration of their services into device scenarios. - Expanded range of voice-enabled features (video playback controls, navigation, etc.). - Added activation phrases grammars and ran usability assessments. - Performed C++/Python code-review, deployed backend releases, wrote tests, fixed bugs.

- Wrote and optimized SQL, SAS, and Python code for marketing automation (MA), credit history upload and integration tasks (both in-bank and external (credit bureau data exchange)). - Developed custom MA platform in Python, which replaced third-party software and cut costs. - Implemented testing framework to verify changes in credit history upload scripts, which helped reduce development and error-correction time, as well as improve customer satisfaction.

Thesis: Method for proving correctness of algorithms using Petri Nets