Dishank Bansal

dishankbansal09@gmail.com

Namaste! Bonjour! I am a Research Engineer at UK AISI working on AI Agents and post-training LLMs. Previously, I was an AI Resident at Meta AI (FAIR) working with Brandon Amos, Ricky Chen and Mustafa Mukadam on reinforcement learning and task-based representation learning. Currently, my work focuses on capability elicitation of LLM agents via post-training and other techniques. My broader interests lie in scalable and generalizable agents.

I completed my Master’s in AI from Mila and University of Montreal, where I worked on robot learning and Bayesian deep learning at Robotics and Embodied AI Lab under the supervision of Prof. Liam Paull.

Before joining as a Master’s student, I was an intern for a year with Prof. Liam Paull. And in what feels like a past life, I was an undergraduate student at IIT Kharagpur, where I spent all of my time at AGV under supervision of Prof. Debashish Chakravarty working on perception and embbeded-control systems for autonomous vehicles.

I am an avid chess and badminton player.

News

Nov 1, 2024 :briefcase: I am joining UK AISI as a Research Engineer in the Agents/Post-training team.
Sep 1, 2023 :page_facing_up: Our TaskMet paper got accepted at NeurIPS 2023.
Jul 1, 2023 :page_facing_up: Our TaskMet paper got accepted at DiffAE Workshop, ICML 2023.
Aug 1, 2022 :briefcase: Started AI Residency at Meta AI (FAIR) working on reinforcement learning.
Jan 1, 2022 :page_facing_up: Our f-cal paper got accepted at ICRA 2022.

Selected Publications

  1. TaskMet: Task-Driven Metric Learning for Model Learning
    Bansal, Dishank, T.Q. Chen, Ricky, Mustafa, Mukadam, and Amos, Brandon
    Advances in Neural Information Processing Systems 2023
  2. f-Cal: Calibrated aleatoric uncertainty estimation from neural networks for robot perception
    Bhatt, Dhaivat, Mani, Kaustubh, Bansal, Dishank, Murthy, Krishna, Lee, Hanju, and Paull, Liam
    IEEE International Conference on Robotics and Automation 2022
  3. Probabilistic Object Detection: Strengths, Weaknesses, Opportunities
    Bhatt, Dhaivat*, Bansal, Dishank*, Gupta, Gunshi*, Lee, Hanju, Jatavallabhula, Krishna Murthy, and Paull, Liam
    AIAD Workshop, ICML 2020