Bilal Hossain

Research

Bilal is a co-author of the counterfactual-rollouts paper. His contributions centre on the confidence-estimation machinery — how should a learned world model represent its confidence in counterfactual rollouts, and what does the confidence have to do for downstream planning to benefit.

He has been with alphabell since 2020 and is based in Dhaka. He is the lab's primary contact for the BUET AI safety reading group.

Background

Ph.D. computer science, BUET (Dhaka), 2014. Postdoc at NYU (2014-2016).

Prior to alphabell: NYU; Volterra Cognition; Wayfarer Institute.

Selected publications

• Dec 2024 · ab-counterfactual
  Counterfactual Rollouts for Planning: a 30-day deployment study
  Sasha Petrov, Maya Quesada, Bilal Hossain
• Aug 2025 · ab-compositional-
  Compositional Latent Dynamics for Long-Horizon World Modelling
  Jonas Bremer, Sasha Petrov, Felicity Anjali Sandirasegaram, Tomoko Niwa
• Jul 2025 · ab-embodied-pretr
  Embodied Pretraining via Cell-Operated Simulation: a 70k-environment study
  Lin Hao, Dimitri Yelchaninov, Sera Wijewardene, Ananya Mukherjee
• Oct 2024 · ab-federated-comp
  A Federated Compute Scheduler for an Asynchronous Research Lab
  Pranav Iyer, Yusra Habibi, Akoss Vidor
• Sep 2024 · ab-interpretabili
  Interpretability Cell Pairing: how every dual-use capability run gets a watchful sibling
  Karima Belkadi, Hester Vandekerckhove, Yuki Cho

Full publications index →

Recent talks

• Confidence and the 30-day study, NeurIPS workshop 2024