Summary
Homin Kwon is a senior computer scientist with a decade of experience building production-grade ASR, KWS, and speech technologies for government and commercial customers from Samsung to SRI International. He leads ASR model development and optimization—specializing in dynamic data augmentation for challenging acoustics, two-pass decoding with LM rescoring, semi-supervised and multilingual transfer learning, and inference-speed improvements (see his faster-wav2vec2 repos). Homin’s background spans embedded mobile voice systems (Bixby/S‑Voice), TTS alignment, and multimodal work like American Sign Language recognition, giving him a rare cross-domain view of speech, video, and language pipelines. He combines rigorous academic training (PhD-level work in electrical engineering) with hands-on production delivery and optimization, including LM pruning, calibration for keyword detection, and practical data curation. Colleagues value his ability to translate research techniques into scalable, resource-efficient systems that perform in real-world noisy conditions.
10 years of coding experience
14 years of employment as a software developer
Ph.D Electrical Engineering, Ph.D Electrical Engineering at Arizona State University