Computational Mathematician - Internship 2025
Fully Remote
Description

 

The future of automation increasingly appears to consist of hybrid teams of humans, foundation models (LLMs, Diffusion Models), and multi-agent semiautonomous onboard agents acting together to achieve a common series of goals. This hybrid environment requires humans to communicate commands and intentions to agents with the potential for discontinuous communication, split-second decision making, and goal reorientation of an entire team of autonomous agents with only minimal input from a commanding user.  


Mobius Logic, Inc. works with partners in US government to understand and identify exploratory AI research that shows sufficient promise and to develop these ideas into practice via Research & Development solutions. Practically, we aggregate Technology Readiness Level (TRL) 2-3 ideas and develop them into TRL 4-5 technologies.   


Recently, we have been developing hierarchical, human-on-the-loop control systems for close proximity satellite operations and are now expanding this research into teams of autonomous agents acting collaboratively with human and non-human systems. We have developed open-ended, skill-based RL training methods that produce fleets of diverse, controllable agents, and are developing the systems and language for humans to effectively interact with these fleets in a semi-autonomous way. There are several projects available to interested interns in this pipeline including: training low level agents using reinforcement learning, developing algorithms for dynamic teaming, course of action generation and human interpretable skill-based planning, and evaluation of multilayered highbred RL/control/OR systems.  


The chosen interns will be part of a team of other members with varied skills in Data Science, NLP, Mathematics, and Software Development. During the internship, the interns will learn and code in Python, integrate their work with the compute platform, make changes to the Mobius Logic DRL engines, attend weekly meetings with project members, read and discuss papers, present their work, and write research reports.     



Requirements
  • Master’s degree or higher in Computer Science, Engineering, Applied Mathematics, Statistics or other technical or mathematical fields.
  • Solid math background. Ability to read a machine learning paper and produce the pseudo code of that paper.   
  • U.S. Citizenship is required.

Desirable Skills:

  • Statistics
  • Deep Learning
  • Control Theory
  • Operations Research or Reinforcement Learning knowledge a plus, but a desire to learn can replace skills in this area. 
  • Python (numpy, stablebaselines3, jax, torch) helpful but not required.