At Cubicalframes, I design and maintain Snowflake pipelines using a Bronze-to-Silver-to-Gold Medallion Architecture, with data ingested from AWS S3. I implemented stream-based incremental loads that cut full-refresh compute costs by about 40% compared with truncate-load.
I also engineered a production speech-to-text pipeline using Whisper, Deepgram, AssemblyAI, and NeMo, improving transcription accuracy by 23% through WER evaluation and automated logging. I build prompt-engineered GPT-4o workflows for structured transcript insights, and have also developed REST APIs and React-based applications.
