Benjamin Scellier

Machine learning research scientist and Principal Research Scientist at Rain AI, working on physics-based computing and physics-based learning algorithms (notably Equilibrium Propagation) aimed at reducing the cost of AI hardware. PhD in deep learning at Mila under Yoshua Bengio; postdoc at ETH Zurich; prior research residencies at Google and X, the moonshot factory. Earlier: MSc Statistics (NUS), MSc Applied Mathematics (Ecole Polytechnique).

Notes6

• Quant employment verification: CONFIRMED. LinkedIn work history (returned via the person index; index-only, self-reported) lists a role as Quantitative Analyst at Prescient Limited, Cape Town, dated 2012, 'Fundamental analysis and portfolio simulations'. Profile identity is corroborated by independent anchors (personal site bscellier.github.io, ORCID 0000-0002-2407-7470 linking the same site, Rain AI current role). Prescient is a South African investment manager whose own site and LinkedIn ('systematic investing') show it operates in quantitative/systematic asset management, not merely a 'quant' name. The role sits between Ecole Polytechnique (2010-2013) and the NUS master's (2013-2015), consistent with a summer analyst stint; role evidence is index-only, so treat employer and fact of role as supported but dates as year-only.2012
• The IMO 2006/2007 France identity exists only in search-result titles and the caller lead; no participant page read in this run yet independently ties the Olympiad contestant to this ML researcher, so it remains lead-only pending a primary check.2007
• Postdoctoral researcher at ETH Zurich 2021-2022, then Principal Research Scientist at Rain AI (2022-present, Zurich/remote) on energy-efficient analog AI processors.2022
• PhD in deep learning at Mila (Quebec AI Institute) 2016-2020 under Yoshua Bengio, who publicly described Scellier as his PhD student; thesis on Equilibrium Propagation.2020
• Author of Equilibrium Propagation, a physics-grounded learning framework; the personal site lists a stream of EP papers (Frontiers 2017, Nature 2025 review, NeurIPS 2023, ICML 2024) - a decade-long single-thread research program.2017
• Research residencies at Google (2019-2021) and X, the moonshot factory (2019); earlier Data Scientist at Knowesis (Singapore, 2014-2015) and Research Intern at University of Oxford (2013, stochastic processes).2013