Roman worked on improving gas fee estimation stability in the osmosis-frontend repository, addressing instability in base fee queries that previously led to unreliable transaction cost predictions. He increased the base fee multiplier from 1.5 to 1.65 across gas fee calculation functions and updated related tests to ensure the new logic was robust. Using TypeScript and JavaScript, Roman focused on aligning frontend calculations with backend fee logic, which reduced estimation failures and improved the overall user experience. His work demonstrated attention to code quality and testing strategies, resulting in higher transaction success rates and reduced support needs for end users.