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.
Hyperconnect is enhancing user experiences across Match Group's diverse brands by leveraging industry-leading AI technology.
Tinder Photo Selector and Hinge Photo Finder are innovative AI based features introduced by the Hyperconnect team in collaboration with Tinder and Hinge. These features assist users in effortlessly selecting optimal profile photos, based on what we’ve learnt as attributes common to good profile shots (e.g. lighting, composition, etc.). Via these features, we simplify the complex process of photo selection, enabling users to express themselves more uniquely and effectively. This, in turn, facilitates deeper and more meaningful connections.
Papers and Awards
- 2024: "A Real-Time Session-Based Reciprocal Recommendation System for a One-on-One Social Discovery Platform", at the ICDM 2024 Workshop
- 2023: “TiDAL: Active Learning Method Based on Model Behavior for Efficient Learning Processes,” at the ICCV 2023
- 2023: “Research on Setting Thresholds to Simultaneously Satisfy Multiple Classification Criteria in Moderation Environments,” at the WSDM 2023
- 2022: “Measuring and Improving Semantic Diversity of Dialogue Generation” at the EMNLP 2022, the best international conference in the field of natural language processing (NLP)
- 2022: “Noisy Label Learning Through Efficient Transition Matrix Estimation to Prevent Label Miscorrection” at the ECCV 2022, the world’s best image processing conference
- 2022: “Meet Your Favorite Character: An Open Domain Chatbot That Imitates a Virtual Character with Simple Speech” at the NAACL 2022, the world’s most prestigious natural language processing conference
- 2022: “Enhancing Dialogue Generation Model Performance Using Examples” at the ACL 2022 Workshop, the top venue for computational linguistics research
- 2022: “Temporal Knowledge Distillation for On-Device Audio Classification” at the ICASSP 2022, the world’s most prestigious speech signal processing conference
- 2021: “Feature Normalization for Importance Preservation in Click-Through Rate Prediction” awarded Best Paper at the ICDM Workshop, a leading conference in data mining research
- 2021: “Efficient Click-Through Rate Prediction Model Based on Tabular Learning” at the ICLR 2021 Workshop, a prominent event for advancements in deep learning research
- 2021: "”Knowledge Distillation Technique from Large-Scale Generative Model to Search Model for Efficient Everyday Conversation"” at the EMNLP 2021, the best international conference in the field of natural language processing (NLP)
- 2020: ‘Long-Tail Image Classification Problem Solution’ at the 2021 CVPR, the world’s best deep learning conference
- 2020: 'Few-Shot Learning for Text-to-Speech (TTS)' published at INTERSPEECH 2020
- 2019: 'Few-Shot Learning for Face Reconstruction' published at AAAI 2020
- 2019:'Fast-Operating Keyword Spotting Mode (TC-ResNet) on Mobile' published at INTERSPEECH 2019
- 2019: “Lightweight Image Segmentation Model (MMNet) Optimized for Mobile Environments” uploaded to arXiv, a preeminent repository for research papers across various fields
- 2018: 2nd place in the LPIRC, Low Power Image Recognition Challenge at 2018 CVPR, a premier annual computer vision and deep learning event
