Software Engineer at Slower -slower.ai

Led the CLI automation tool to optimize geocoding engine builds, processing billions of geospatial records with BigQuery data warehousing and cutting build time from 90 days to 53 days, with ongoing optimizations targeting 30-day builds. Optimized Apache Airflow DAG processing on GCP through task parallelization, code refactoring, and infrastructure tuning, achieving a 97% reduction in data processing time (from 4 days to 2.5 hours). Enhanced production monitoring and alerting by configuring Elastic Kibana dashboards and PagerDuty rules, improving reliability and incident response. Integrated the company’s general-purpose AI assistant into the tax filing platform as an interview-style agent (FastAPI, TypeScript) and built its Python tool layer for form extraction, prior-year return lookup, eligibility checks, and deduction recommendations as discrete MCP endpoints used by the LLM agent during interviews. Adapted the hybrid RAG knowledge base to the tax domain over Aurora PostgreSQL b