• Built and maintained large-scale data pipelines supporting Siri data processing and validation workflows.
• Designed an ML-powered audio pre-annotation pipeline reducing manual labeling effort by ~50%.
• Built org-wide plugins for LLM-assisted Coding Agents to enhance engineering productivity.

Tech stackPythonPandasPydanticClaudeMCPStreamlitSparkAWS

• Worked on the Enterprise Document Parser evaluation tool using Python ecosystem that focuses on enhancing the accuracy of structure and content recognition for parsing various complex documents, automating workflows, and external services integration.
• Re-architected the core scoring algorithms of the evaluation tool, dedicated to capture edge cases that ultimately transform metrics into actionable insights.
• Automated regression analysis for parser evaluation, cutting the manual processing time from 4+ hours to less than 10 minutes, boosting team productivity and saving 60+ engineering hours weekly.
• Engineered scalable backend solution by implementing a multiprocessing module to process over 500+ documents in parallel, reducing execution time by 75% and ensuring seamless pipeline integration.
• Built the entire CI/CD pipeline for code quality assurance, and set up a weekly cron job for evaluating and processing a massive volume of files which enable hands-free experience for everyone.
• Developed a reusable cloud storage client that integrated into 3+ projects, standardizing the cloud storage interactions and reducing redundant development effort by 40%.

Tech stackPythonPandasPydanticOpenXMLCI/CDCloud SDKAutomation