ASR Evaluation & Model Leaderboard
Testing ASR models in an early-stage speech-recognition pipeline, and keeping a shared leaderboard that helps the team compare and select models.
- Test ASR models on datasets I sourced, flagging pipeline issues and correcting a model setting that improved transcription accuracy
- Took ownership of an agentic LLM workflow that adapts transcripts for ASR use, learning the system and demoing it to the team
- Built and maintain an interactive leaderboard web server that ranks tested ASR models by capability