Amro Abbas
Datologyai 
đź“Š It's always a multi-step, data-driven, literature-cited investigation.
🖥 If it can be parallelized on a GPU cluster, that’s art.
I am a founding member of the technical staff at DatologyAI, where I spend my time conducting research aimed at understanding how deep models learn from data. Prior to that, I was an AI Resident at Meta AI working with Ari Morcos. I hold a masters’ in Math and Computer Science from the African Institute For Mathematical Sciences (AIMS) (the African Master’s in Machine Intelligence (AMMI)).
Do not hesitate to reach out for collabiration.
news
| Mar 01, 2024 | Joined DatologyAI as a founding member of technical staff. |
|---|---|
| Jan 01, 2023 | Our paper (SemDeDup: Data-efficient learning at web-scale through semantic deduplication) received the Best Paper Award @ The Multimodal Representation Learning Workshop at ICLR 2023. |
| Nov 01, 2022 | New paper accepted at AAAI2023 (Progress and limitations of deep networks to recognize objects in unusual poses). |
| Sep 01, 2022 | I joined Meta AI in California as an AI Resident. I will be working with Ari Morcos’s team. |
| Oct 01, 2020 | I completed my BSc at University of Khartoum / Electrical Engineering (major in software engineering) |
selected publications
- Semdedup: Data-efficient learning at web-scale through semantic deduplicationBest Paper Award @ The Multimodal Representation Learning Workshop, at ICLR 2023,
- Datacomp-lm: In search of the next generation of training sets for language modelsNeurIPS 2024,
- Effective pruning of web-scale datasets based on complexity of concept clustersICLR 2024 Oral Presentation @ DataComp Workshop at ICCV, 2023 ,
- Progress and limitations of deep networks to recognize objects in unusual posesAAAI 2023,
- Sieve: Multimodal dataset pruning using image captioning modelsICLR 2024,
- A comparison between humans and AI at recognizing objects in unusual posesTMLR 2024,