Ostap Hamarnyk

Designed and built an autonomous agent to assist with the production support lifecycle for enterprise RPA bots. To address a high volume of L1 support tickets due to transient bot failures, I architected a solution that interfaces with OpenAI's API to analyze error logs in real time.

I led the development of the Python backend, which handled the LLM context management, ensuring the model could accurately classify errors as "system," "application," or "data" exceptions. The agent was then programmed in UiPath to trigger specific self-healing workflows based on this classification, automatically restarting hung applications or sending detailed, pre-analyzed reports to human developers for complex code issues.

This project transformed our support model from reactive to proactive, reducing the mean time to resolution (MTTR) by over 60% and allowing the core development team to focus on new features rather than maintenance.

Took ownership of an underperforming invoice processing automation to drive higher efficiency and accuracy. I performed a comprehensive code refactoring of the legacy solution to improve modularity and maintainability.

I optimized the UiPath Document Understanding model by retraining extractors to better handle variable invoice layouts and reduce data-extraction errors. Crucially, I engineered advanced post-processing logic to automatically validate extracted data against ERP records. This new logic intelligently filtered out non-critical discrepancies and "noise," preventing unnecessary exception handling for the accounting team.

These optimizations drastically reduced manual touchpoints, resulting in a 30% increase in straight-through processing (STP) and allowing the finance team to trust the automation for high-volume transactions.