i like cli

Nyx Iskandar

GitHubLinkedInORCIDScholar

Hey! I'm Nyx.

I'm currently leading research at Ramen VR / Aura as the first member of Ramen Labs. I'm also a fellow at V11. I spend my days reading, thinking, musing, writing, coding, climbing, and meeting and learning from the coolest people.

I graduated from UC Berkeley EECS (Honors) in 2.5 years. I was active as an undergrad ML researcher at BAIR and CHAI, as well as VP of Machine Learning at Berkeley. After the AI lab, my local bouldering gym was (and still is) my next most frequented place.

I've learned that I love solving the hardest puzzles in ML alongside the smartest and most driven of them all. I take risks and seek new challenges; falling off walls teaches you more than reaching the top, though only if you aim to reach the top, and only if you pick harder and harder walls to climb.

RESEARCH

Curr: 3D unified multimodal models

Prev:

  • Action abstractions for efficient planners (with Dr. Cam Allen)
  • Language model pre-training, evals, and wargame simulations (with Dr. Ritwik Gupta)
  • Inverse RL (as OpenAI Research Fellow)
  • LLM agent tool usage strategies for concise context (as Foam Founding Engineer)
  • Emergent symbols/language for communicating abstract concepts (pet project)

I've published workshop papers at NeurIPS and ICML, and contributed to a chapter in a technical book. I plan to continue publishing.

INDUSTRY

Curr: Research Engineer @ Ramen VR

Prev:

  • Founding Engineer @ Foam
  • Research Intern @ GovTech Singapore
  • AI Engineer Intern @ Ramen VR

SELECTED PUBLICATIONS

  1. Eta Given Delta: Defining LLM Tool Efficiency With Marginal Tool Utility [Preprint]
  2. A Matter of Representation: Towards Graph-Based Abstract Code Generation [NeurIPS 2025 DL4C Workshop]
  3. Measuring what matters: A framework for evaluating safety risks in real-world LLM applications [ICML 2025 TAIG Workshop]

Thanks for stopping by!