AI Lab MLAI Lab ML(Machine Learning) , , . // , . , , ML Platform end-to-end .
AI Lab AI . AI / . SotA . AI flywheel .
AI lab . AI lab AI , . AI lab .
• • x ML Platform Team• x ML Team
AI Lab AI Lab . , . . , AI Lab .
• 2018 (LPIRC) 2• 2019 (MMNet) • 2019 (TC-ResNet) INTERSPEECH 2019 • 2019 AAAI 2020 • 2020 Text-to-Speech(TTS) INTERSPEECH 2020 • 2020Long-tailed Visual Recognition Label distribution shift
. / .
. ML Platform . . .
[Content Understanding] . .
• • label multi-task multi-label • Partial multi-modal data
[User Understanding]"Connect people through shared interests and create enjoyment" AI Lab , . .
• content understanding (ex. / ) • cold-start • (reciprocal)
[ML in Production] Kaggle . .
• Highly imbalanced or noisy label data • Active learning core-set selection • deploy continual/life-long learning • task meta-learning • privacy federated learning
[Open-domain Conversational Agent] AI . AI Lab conversational agent chatbot system , .
• Conversational agent • Large scale language model (over billion parameters) , optimization, distillation • Conversational agent • Human-in-the-loop conversational agent • Conversational agent
• on-premise .• 1• , Kubeflow pipeline BigQuery ML Platform AI lab .
/
., . , .
Work (8am - 11am) , 2 9 , , , 18 ,
Life (Hypresso) , , ! (, , )/( 9 / )/() ( ) ( ) ( ) 3 / , / 30
Growth amp; Refresh , , / / , , , , , , IT 3 (10) (200) ( 20) + , (//)