1. Deep model optimization: We design a holistic framework for embedded AI models, which includes data dimension reduction, few-shot learning, model compression etc, to build lightweight models for a wide range of embedded platforms.
2. Real-time inference: We develop a real-time scheduling framework which can support 10+ real-time deep learning tasks to run concurrently via CPU-GPU co-optimization, while 3x faster than the SOTA.
3. On-device learning: We design a novel cluster-based federated learning framework, which can keep data local while training. Therefore, this on-device learning paradigm can protect user’s privacy. Also, it can be deployed for large-scale IoT systems.
4. Novel algorithms for next-generation sensors. Next-Gen sensors require novel algorithms for specific data processing and applications. We have 10+ years of experience in developing advanced algorithms for next-gen sensors such as depth sensors, mmWave radars, and LiDARs.
如何解決市場痛點 | RESPONSE TO MARKET PAIN POINTS:
1. Our deep model optimization technologies can Minimize resource usage of deep learning models that can be trained with minimal data.
2. Our capability of real-time inference can enable accurate inference within bounded delay.
3. Our on-device learning technologies can preserve users’ data privacy.
現時或曾經參與之創業培育計劃 ｜ INCUBATION PROGRAMME:
1. Technology and Business Development Fund (CUHK): HKD 200,000
2. Incu-Tech (HKSTP): HKD1,290,000
3. Technology Start-up Support Scheme for Universities (HK gov.): HKD 600,000
現在尋找… | LOOKING FOR…:
資金 Funding 創業夥伴 Partners 曝光機會 Visibility and Exposure