Eylam Tagor

Bio

I am a PhD candidate in Computer Science at Brown University, advised by the fantastic Prof. Serena Booth. My research focuses on reinforcement learning and causal inference, particularly in building reliable decision-making systems that can handle uncertainty, partial observability, and real-world imperfections. I believe that advancing causal reasoning capabilities is essential to improving human-robot collaboration.

I completed my M.S. at Columbia University advised by Prof. Elias Bareinboim and working on causal RL and imitation learning, and my undergraduate degree at UT Austin where I worked with Prof. Peter Stone on multi-agent RL and sim2real as part of the RoboCup team.

Research

Tagor, E., Li, M., & Bareinboim, E. Scalable Causal Imitation Learning: Introducing new causal algorithms for confounding robust imitation learning in high dimensional, long horizon continuous control tasks.
RLC 2026 · PDF · Code

Gymnasium for Causal Imitation Learning: Developed imitation learning algorithms grounded in causal inference; built SCM-parameterized RL environments supporting observational, interventional, and counterfactual evaluation.
2025 · PDF

Residual Neural Causal Model: Designed ResNCM, a neural causal model using PyTorch and Causal-Learn; conducted counterfactual analyses on education datasets, outperforming classical and deep-learning baselines.
2025 · PDF · Code

Multi-Agent RL for RoboCup @ LARG: Implemented multi-agent deep RL behaviors for NAO robot soccer; improved sim2real transfer using high-fidelity simulation, contributing to a first-place finish in the 2024 RoboCup SPL Challenge Division.
2024 · Code

Generalist Embodied Video Game Agent @ NVIDIA: Built embodied agents and multimodal training environments using OpenAI Gym; integrated RL algorithms, LLMs, and large-scale pretraining to improve cross-task generalization.
2023 · Code

Teaching

• ORCS 4529: Reinforcement Learning (Columbia, Fall 2025)
• COMS 4995: Causal Inference for Data Science (Columbia, Spring 2025)