Sergey Yorsh

Developed backend services for a multi-country payroll processing platform, focusing on UK and German compliance rules and regulatory reporting. Worked across the full service layer: business logic, database access, API design and background processing.

Part-time contract. Backend for an iOS app that identifies coins from a photo and estimates their value. Took over a legacy serverless backend and made it stable under paid traffic. Built a large-scale data collection pipeline for the coin catalog. Implemented image recognition with vector similarity search and price aggregation from multiple sources, with estimation of missing grade prices. Set up error monitoring with Sentry and reduced Lambda latency.

Developed WiFi and Bluetooth interface services in a Linux environment. Implemented a data collection service in Rust. Worked with the MQTT protocol for device communication. Contributed to embedded systems development using Yocto Project and MicroPython for IoT devices.

Developed data scraping and cleaning platform for neural network training. Led migration from monolith to microservices architecture while maintaining database integrity. Created AlchQl library enabling seamless transition to Apollo Federation and solving N+1 query problem for SQLAlchemy queries. Improved system performance and maintainability through service decomposition. Designed and maintained data scraping and cleaning pipeline to produce high-quality training datasets for price prediction ML models.

Built marketing analytics platform from scratch. Designed and implemented microservice architecture for real-time ad performance analysis. Developed robust ETL pipelines using Celery and RQ for processing large datasets. Implemented high-load data processing systems with MongoDB and PostgreSQL.

Developed backend infrastructure for social goal-tracking platform. Built API using Odoo framework with Python 2.7 and Angular 1 frontend.

Was one of the founding developers from startup stage. Initially developed web scraping infrastructure. Later led a dedicated data collection team while developing Django backend. Implemented distributed scraping architecture and data quality control systems. Applied NLP techniques (using Natasha) for extraction and normalization of structured temporal data from unstructured text sources.

Was involved in the development of devices for Quest rooms. Arduino with various modules was most often used. Gained extensive experience working with hardware and knowledge of low-level communication protocols.