We develop a small and lightweight deep learning model optimized for mobile devices.
Hyperconnect seeks to better understand users without compromising data security and privacy. We ensure that our machine learning models run on-device without user data ever having to leave the device
We make easy interaction between users possible.
Our model learns from every user interaction. The more it learns, the better it becomes at providing more engaging and authentic experience for users.
We develop and directly design the AI infrastructure, along with research and production engines.
Based on a solid ML-Ops culture, we strive to achieve smarter and faster research iterations. We are automating everything, from smart data processing to cutting-edge model research and development.
Papers and Awards
- 2018, Low-Power Image Recognition Challenge (2nd place)
- 2019, Towards Real-Time Automatic Portrait Matting on Mobile Devices (Arixv)
- 2019, Temporal Convolution for Real-time Keyword Spotting on Mobile Devices (INTERSPEECH19)
- 2019, MarioNETte: Few-shot Face Reenactment Preserving Identity of Unseen Targets (AAAI20)
- 2020, Attentron: Few-shot Text-to-Speech Exploiting Attention-based Variable Length Embedding (INTERSPEECH20)
- 2020, Disentangling Label Distribution for Long-tailed Visual Recognition (CVPR21)
- 2021, Distilling the Knowledge of Large-scale Generative Models into Retrieval Models for Efficient Open-domain Conversation (EMNLP21)
- 2021, Efficient Click-Through Rate Prediction for Developing Countries via Tabular Learning (ICLR21 Workshop)
- 2021, Embedding Normalization: Significance Preserving Feature Normalization for Click-Through Rate Prediction (ICDM21 Workshop)
- 2022, Temporal Knowledge Distillation for On-device Audio Classification (ICASSP22)
- 2022, Meet Your Favorite Character: Open-domain Chatbot Mimicking Fictional Characters with only a Few Utterances (NAACL22)
- 2022, Learning with Noisy Labels by Efficient Transition Matrix Estimation to Combat Label Miscorrection (ECCV22)
- 2022, Measuring and Improving Semantic Diversity of Dialogue Generation (EMNLP22)
- 2022, Reliable Decision from Multiple Subtasks through Threshold Optimization: Content Moderation in the Wild (WSDM23)