Ahmed Medhat

Summary

Over the past 10 years, I have been devoted to network science and graph learning research within both academia and industry. Working as a research scientist in network science groups at Oxford and Facebook, building a data startup that today employs over 100 people and leading diverse teams of research scientists and engineers. I conducted and published fundamental and applied research, which, amongst other uses, has been used by the UK prime minister's office in public debates, informed how social network companies tackle misinformation and hate speech, and how NGOs such as WHO and UNHCR responded to the COVID pandemic and the Ukrainian refugee crisis, respectively. I have also acquired extensive experience in the end-to-end process for building predictive ML models, especially in creating algorithms and systems for ground truth generation of visual and natural language labels, and learning from weakly supervised data, through managing a large scale, multi-million dollar budget labeling project involving 100s of labelers labeling content across 10s of languages.

I am currently a senior research scientist in Meta's Networks & Behavior Modeling Team, part of the Computational Social Science Group. My work includes both fundamental and applied research at the intersection of network science, graph learning, behavioral science and social computing. My recent research involved adopting a variant of bootstrap percolation to study how structural biases in social networks, particularly expressed via the friendship paradox, can shape people’s propensity to share content on newsfeed based social media platforms (article pre-print).

Over my tenure at Facebook, I have worked on numerous machine learning and network science projects, including building models to improve conversational health, implementing scalable multi-layer centrality and node attribute prediction methods using GNNs, improving the way we measure network effects and utilizing human geographic co-location patterns for epidemiological modeling for the COVID pandemic (article pre-print). I also have extensive experience in implementing algorithms and systems for ground-truth generation of natural language content labels, and in generating and learning from weakly supervised data.

While part of Meta’s Data Science for Social Good team, I was involved in creating global scale network and geographic datasets, particularly colocation maps and displacement maps, that have aided WHO and the UNHCR in response efforts to the COVID pandemic and the Ukrainian refugee crisis, respectively (press coverage).

In my free time, I'm an avid pool and snooker player. (shameless plug)

Website powered by w3.css, sidebar image courtesy of the Network Neuroscience Journal