During August 2025, Marek contributed to the softwaremill/sttp-ai repository by developing a feature that exposed an optional usage field in chat completion streaming responses. He extended the backend data model using Scala and JSON to include detailed token counts for prompts, completions, and totals, enabling improved analytics and cost estimation for streaming chat usage. Marek implemented comprehensive unit and integration tests to ensure accurate deserialization of usage data, focusing on robust API integration and data modeling. His work enhanced observability and maintained streaming performance, demonstrating depth in backend development and a methodical approach to extending functionality without introducing regressions.