Asim Kadav
Research Interests: Content understanding and generation and efficient ML systems
News: Watch my lecture on Video understanding at Yale University, Topics in Deep Learning, Spring 2020.
Asim Kadav is an experienced technical leader with over 15 years in developing and deploying cloud and edge-based AI products and large-scale distributed systems. He has led multiple successful product launches and is a researcher with 8000+ research citations and 50+ patent filings. Most recently, Asim led teams at Samsung Research America, Freenome, and Tonal, where he focused on developing sub-1B parameter Vision-Language Models for video understanding and planning, ML models for early cancer detection and GenAI biomedical and clinical applications, and launched Tonal's first camera-based computer vision product. Before that, he led the eigen team at NEC Labs, advancing efficient image and video classification, retrieval, and pose tracking methods. With a strong background in multi-modal learning and distributed systems, he has developed and deployed several scalable video AI products. Asim's past work extends to building data center solutions, distributed file systems, and OS reliability tools, with publications in top conferences such as ICLR, CVPR, ECCV, SOSP, OSDI, and ASPLOS.
Email: firstlastname@gmail.com
Recent Publications | All, By area | Students
- COMPOSER: Compositional Reasoning of Group Activity in Videos.
Honglu Zhou, Asim Kadav, Aviv Shamsian, Shijie Geng, Farley Lai, Long Zhao, Ting Liu, Mubbasir Kapadia, Hans Peter Graf.
In ECCV 2022. PDF.
- Hopper: Multi-hop Transformer for Spatiotemporal Reasoning.
Honglu Zhou, Asim Kadav, Farley Lai, Alexandru Niculescu-Mizil, Martin Renqiang Min, Mubbasir Kapadia, Hans Peter Graf.
In ICLR 2021. PDF. Poster.
- 15 Keypoints is All You Need. Michael Snower, Asim Kadav, Farley Lai, Hans Peter Graf. In CVPR 2020.PDF. Ranked #1 in PoseTrack.
- S3VAE: Self-Supervised Sequential VAE for Representation Disentanglement
and Data Generation.
Yizhe Zhu, Martin Renqiang Min, Asim Kadav, Hans Peter Graf.
In CVPR 2020. PDF.
- Tripping Through Time: Efficient Localization of Activities in Videos.
(Spotlight)
Meera Hahn, Asim Kadav, James M. Rehg, Hans Peter Graf.
In CVPR Workship on Language and Vision, 2019. Also, appears in BMVC'20. PDF.
-
Visual Entailment: A Novel Task for Fine-Grained Image Understanding.
Ning Xie, Farley Lai, Derek Doran, Asim Kadav.
In NeurIPS Workshop on Visually-Grounded Interaction and Language, 2018 (ViGIL’18). PDF.Dataset. Check out the leaderboard!
- Teaching Syntax using Adverserial Distraction. Juho Kim, Christopher Malon, and Asim Kadav. In FEVER-EMNLP, 2018. PDF.
- Attend and Interact: Higher-Order Object Interactions for Video Understanding. Chih-Yao Ma, Asim Kadav, Iain Melvin, Zsolt Kira, Ghassan AlRegib, and Hans Peter Graf. In CVPR, 2018. PDF. Video NEC Video. Live Demo.
- Adaptive Memory Networks. Daniel Li, Asim Kadav. Accepted at ICLR '18 (Workshop Track).
PDF.
- A Context-Aware Attention Network for Interactive QA. Huayu Li, Martin Renqiang Min, Yong Ge, Asim Kadav. In KDD, 2017.
PDF. Poster.
- Pruning Filters for Efficient ConvNets. Hao Li, Asim Kadav, Igor Durdanovic, Hanan Samet, Hans Peter Graf. In ICLR'17.
PDF. Poster. Code (Reproduced by others).
- DeepConfig: Automating Data Center Network Topologies Management with Machine Learning. Christopher Streiffer, Huan Chen, Theophilus Benson, Asim Kadav. SIGCOMM Net AI Workshop, 2018. PDF.
- ASAP: Asynchronous Approximate Data-Parallel Computation.
Asim Kadav and Erik Kruus. arXiv Preprint. Dec 2016.PDF . Code.
- Privacy preserving collaboration in Bring Your Own Apps. Sangmin Lee, Deepak Goel, Edmund L. Wong, Asim Kadav, Mike Dahlin. In Symposium on Cloud Computing (SOCC '16).
PDF.
- Random Walk Distributed Dual Averaging Method For Decentralized Consensus Optimization.
Cun Mu, Asim Kadav, Erik Kruus, Donald Goldfarb, and Martin Renqiang Min.
NIPS Optimization Workshop (NIPS OPT'15), Montreal, Canada, December 2015.
PDF .Extended PDF .Poster.
-
MALT: Distributed Data-Parallelism for existing ML applications.
Hao Li, Asim Kadav, Erik Kruus, Cristian Ungureanu.
ACM European Conference on Computer Systems, (Eurosys '15) , Bordeaux, France, April 2015.PDF .Poster. Slides
Open-source software/datasets
- CATER-H: A more difficult CATER dataset with longer reasoning chains, ICLR 2021. Github.
- SNLI-VE: A novel visual entailment task using image/text pairs. Github.
-
MALT-v2: A federated learning framework over RDMA and TCP. Github. Website.
PDF . - Numerous patches to the Linux and android kernel to improve the software/hardware layer.[1][2][3][4].
Press
- Are You Making Form Mistakes When Lifting? Tonal Smart View Can Show You.
- NEC Corporation of America is Partnering with Haven for Hope to Co-create Technology-Enabled Solutions That Enhance Safety, Streamline Operations and Empower Clients
- Trusting the Hardware Too Much
Service: (out of date)
- Program Committee: AAAI 2019, SoCC 2018, ICML 2018, USENIX ATC 2018, AAAI 2018, SoCC 2017 (Scholarship Chair), ICML 2017, APSys 2016.
- Conference Reviewer: CVPR 2020, ICLR 2020, ICCV 2019, CVPR 2019, ATC 2019, ICLR 2019, ASPLOS 2019 (ERC), NIPS 2018, ICLR 2018(+W), NIPS 2017, NIPS 2016, USENIX ATC 2011, SIGMETRICS 2010.
- Journal/Workshop: NIPS LearningSys 2019, NIPS LearningSys 2018, NIPS LearningSys 2017, NIPS LearningSys 2016, EMDNN 2016, IEEE Computing, Software; ACM: TACO, TECS, TOCS.