Scholar
Marco Cusumano-Towner
Google Scholar ID: VYiRfCwAAAAJ
Apple
Citations & Impact
Citations
972
H-index
12
i10-index
15
Publications
20
Co-authors
28
list available
Publications
Resume (English only)
Academic Achievements
- Selected papers: 'Reinforcement Learning for Long-Horizon Interactive LLM Agents' (arXiv, 2025), 'Robust Autonomy Emerges from Self-Play' (ICML 2025), 'Recursive Monte Carlo and Variational Inference with Auxiliary Variables' (UAI 2022), 'Interval Estimators of Entropy and Information Measures via Inference in Probabilistic Models' (AISTATS 2022), '3DP3: 3D Scene Perception via Probabilistic Programming' (NeurIPS 2021), etc.
Research Experience
- Currently a research scientist at Apple in Vladlen Koltun's research org. Previously, a technical lead at an early-stage molecular diagnostics startup backed by Sequoia Capital.
Education
- PhD in EECS at MIT, advised by Vikash Mansinghka and Josh Tenenbaum; MS in Computer Science at Stanford, researched machine learning for genomics; BS in EECS at UC Berkeley, worked with Pieter Abbeel on household robotics.
Background
- Research topics include reinforcement learning for LLM agents and multi-agent deep RL for autonomous driving. PhD research focused on generative models that include stochastic structure and black box code execution, probabilistic inference in these models (e.g., sequential Monte Carlo, variational), and the compositionality of inference processes.
Miscellany
- Academic research has been funded by the NSF GRFP and the NDSEG fellowship.
