Hi, I'm Pedro, an AI researcher working on ASR and object detection in Goiânia

I'm Pedro Bittencourt, an AI researcher at the Federal University of Goiás and CEIA in Goiânia, Brazil. My work focuses on speech recognition (ASR) and object detection models. On Hugging Face I maintain the Open ASR Leaderboard PT-BR, a Space for comparing speech recognition benchmarks in Brazilian Portuguese, along with a few public datasets behind it, including a cleaned-transcription collection of about 8 million rows.

I'm still early in sharing this work publicly, and I have no public models yet. What I'm figuring out is how to make Portuguese ASR evaluation more useful and trustworthy for other builders.

I'd love feedback from anyone who has worked with ASR in Portuguese or another lower-resource language: which benchmarks or datasets would you want to see on the leaderboard?