I have been head of software development department ever since the creation of the company. The primary product is an application that uses speech recognition algorithms to automatically analyze and to give insights into a huge audio database of gathered call recordings for any call center or sales department of any business (we primarily target b2b sector). We automate and cover a whole range of quality assurance processes by analyzing terabytes of audio data to assess individual call center operators. I interviewed and hired the development team (up to 5 programmers at different times), established and automated development, testing and deployment processes by using Gitlab pipelines and enforcing strict linting rules and code conventions. Designed from ground up project architecture and technical infrastructure, making it scalable from day one, which turned out to be essential in the future (now it's 30 dedicated servers in production). I also wrote most of back-end code base in Python (Flask/SQLAlchemy/numpy). The software involves heavy usage of data science methods, such as clustering, classification, semantic analysis, speech recognition and voice activity detection algorithms. The company has thus become a stable revenue generating business, aimed at b2b sector, since 2019. The initial service architecture has shown its robustness over time. The development department now has transparent internal processes and is easily scalable for arbitrary workloads, which is a huge asset and a strong point for a company.

The task was to develop domain specific engineering application featuring CPU-intensive statistical analysis and a web interface for internal use by aerospace engineers at SAFRAN. In-house mechanical engineering team lacked essential software development skills and practices so I was hired to design distributed application architecture and lead the development process. Right from the start I integrated Gitlab CI/CD pipelines and automated code review, linting, testing and building stages, which made the development process scalable and transparent. As a result, our team satisfied all the deadlines and created a well-documented, modular and easily modifiable application on top of Dash framework, which allowed to implement a UI rich frontend without a single line of Javascript — only Python.Tech stack: GitLab CI/CD, Linux, Docker, Python, Dash framework, scipy, matplotlib, Sphinx documentation.

Being a part of a small team, we developed a platform for auto-testing graphical interfaces of video surveillance software AxxonNext. The objective was to automate manual labor of QA engineers and potentially substitute their work by the software entirely. I also created single-handedly a specialized testing module for the Moscow Surveillance System that automated repetitive test scenarios and eventually promoted the system successful incorporation into the city infrastructure.