Han Teakjin

1. Data Acquisition and Utilization

• Acquired 2023 NIA dataset and built 2024 dataset for large-scale model training.

• Designed datasets with labeled speech data, ITN, and Context Biasing LM integration.

2. End-to-End ASR Algorithm Development

• Developed E2E speech recognition models, including Multi Head Attention expansion and Context Biasing LM.

• Built lightweight models for low-spec devices (2% CPU usage reduction, 80M memory reduction).

3. Android App Development and Service Stabilization

• Developed voice phishing detection sample webpage and Android ASR applications.

• Designed and documented APIs, collaborating with cross-functional teams to meet requirements and optimize performance.

4. Experienced in creating technical content for blogs and social media