I'm Yaroslav Gneushev, a Python backend engineer with 4+ years in fintech and insurance. I build APIs and event-driven systems and keep them running under load — from data models to monitoring and incident response.
Commercial projects: my contributions, technical decisions and results.
The service checks invoices from 12,000 clinics and labs before payment. I implemented antifraud rules based on Ministry of Health clinical guidelines, built an ML scoring microservice in FastAPI and integrated it into invoice processing.
I built a LangGraph agent with MCP servers and integrated it into CI/CD. It reviews merge requests for style, common defects and missing tests before human review. Code stays inside the corporate environment.
Reports lagged by hours and competed with transaction processing in PostgreSQL. I designed a Kafka analytics pipeline with a transactional outbox: events are stored alongside business data, then published to the queue. Data lag fell to tens of seconds.
I consolidated government registry records and PDFs into a common PostgreSQL schema and built APIs and dashboards with route graphs. The parser handles rate limits, retries and corrupted files.
I divide systems into services with explicit boundaries and typed contracts. I choose FastAPI for asynchronous integrations and Django/DRF for complex business logic that benefits from an ORM and admin interface. I verify critical scenarios with integration tests against real infrastructure.
I use PostgreSQL as the primary transactional store, ClickHouse for analytics, and Redis for caching and temporary data. Celery handles background tasks, while Kafka carries events. A transactional outbox keeps data changes and events consistent.
I containerize services and deploy them to Kubernetes through GitLab CI. I monitor system health in Prometheus and Grafana, covering both technical and business metrics. Worker counts and resource limits are based on load test results.
I isolate models behind a versioned API so they can be updated independently of the main application. I build agents as explicit graphs of steps in LangGraph, limiting iterations, controlling tool calls and keeping an execution log.
SO
Built health insurance antifraud services and internal AI tools. I was responsible for design, development, deployment, monitoring and on-call support. Moving heavy operations to priority queues and rewriting problematic ORM queries reduced peak queue depth by up to a factor of 10. After I set up monitoring, the time from incident onset to detection fell from roughly 60 minutes to 5 minutes.
Ap
Worked on the payment platform core: a merchant-facing REST API and transaction processing. I moved analytics to Kafka through a transactional outbox to reduce contention with PostgreSQL transaction processing. Data lag fell from hours to tens of seconds. SQL and index optimization reduced REST API p95 from 2.5 to 0.8 seconds. I also built a payment status notification service and moved processing to an asynchronous workflow.
Ci
Built backend services and analytics for a freight monitoring system, along with ETL pipelines for analytical models. I normalized 100,000 product listings from B2B catalogs to a common schema. When labeled training data was insufficient for OCR, I built a synthetic document generator. In a separate project, I built an ETL pipeline that processed more than 1 TB of text.
Studying Applied Informatics at the Faculty of Information Technology and Big Data Analysis, with a focus on information systems in economics and finance. I study alongside my work.
Studied Programming in Computer Systems and qualified as a software technician. I started using Python during college and took my first engineering job while studying.
