News and Research

OmniML demonstrated significant gains in model performance and cost reduction in its initial collaborations with large enterprise customers in multiple vertical markets, with ML tasks running 10 times faster on different edge devices and 50% in time savings

MCUNet [NeurIPS’20 spotlight][NeurIPS’21]:

SpAtten [HPCA’21]

DiffAugment [NeurIPS’20]

Once-For-All Network [ICLR’20]

Temporal Shift Module [ICCV’19]

ProxylessNAS [ICLR’19]