Next-Gen Materials and Tools for Neuroscience Research, Aberdeen Proving Ground, Md.
Several ongoing and future projects at ARL require the use of so-called “surrogate” materials that are used in place of biological specimens or human subjects for testing the efficacy and effects of different kinds of sensing devices. This project involves the design and investigation of the mechanical, electrical, biomagnetic, and functional properties of various materials and model tools that will be used for neuroscientific study of neurobiological substrates such as brain, bone, or skin tissue. The topic is very multidisciplinary and bridges areas of neuroscience, biomedical engineering, materials science and biochemistry to yield a proper understanding of how novel materials and designs can be used as replacements for live biospecimens in neuroscience research.
Work will be largely hands-on and lab based, taking place at Aberdeen Proving Ground with potential for travel to partner sites, but with the possibility of occasional telework as appropriate. Example duties may include construction and characterization of materials simulating brain, bone, or skin using lab techniques; investigation of various materials and techniques for appropriateness; design and developing physical models and/or techniques for constructing physical models; use of CAD for model development; or use of additive manufacturing for molds, models, or components. Initial work detail is presumed to be a summer or semester length with the possibility of extension as deemed appropriate.
Deliverables include a summary presentation in oral or poster format.
Data Management and Learning for Autonomous System, Middle River, Md.
High speed robots, such as the DARPA Robotic Autonomy in Complex Environments with Resiliency (RACER) Robotic Fleet Vehicle, require immensely higher resolution and higher frame rate sensors to collect data at greater distances which means the data collected by robotic platforms increases in veracity and volume requiring the high-performance computing resources for analysis and machine learning. At any given DARPA RACER test, upwards of 45 TB are collected and must be uploaded on an high performance computing platform for post processing. The ARL autonomy stack is a collection of continuously evolving perception, planning, control and state-estimation algorithms that enable autonomous maneuver. These tools are created in collaboration with other efforts such as the ARL Scalable, Adaptive, and Resilient Autonomy (SARA) Collaborative Research Alliance and DARPA RACER, allowing researchers to process and inspect sensor logs at the terabyte scale and create smaller extractions from these logs for closer analysis and in-turn improving the autonomy stack. Working with the ARL team, the intern(s) will upload data to HPC for processing and labeling. The intern will learn how to train models using these datasets.
All majors are welcome, but experience with a Linux environment, command line and Python3 would contribute to the success of the intern.
Assessing Safety of Frontier Large Language Models in the Energetics Domain, Indian Head, Md.
ETC currently has various projects concerning the safety and accessibility of frontier large language models (LLMs) such as ChatGPT, Claude and Llama. These models often undergo safety checks for various types of potential threats, such as those related nuclear or biological weapons. However, limited work has been done to establish a means to investigate the safety of these tools in regard to energetic systems, such as explosive. ETC is working to study the types of information contained in these models and to develop means to improve model safety in the future.
Interns would work on developing data entry pipelines, establishing tests to evaluate LLM safety, incorporating human feedback into algorithmic tests, as well as other tasks. Work will be fully remote. While any majors are welcome to apply, a background in chemistry, physics, computer science, or any other science/engineering is preferred. Linux/Unix experience and programing experience in Python or C/C++ is preferred but not required.
Machine Learning for Chemistry Application, Indian Head, Md.
ETC has been at the forefront of using machine learning (ML) for the prediction of chemical properties and generation of novel energetic materials. With the rapid improvement of multi-modal large language models (LLMs), ETC is working to incorporate chemical information with conventional chemical texts. This involves developing novel means to describe chemicals and their properties in a manner which can be fused into large language models for improved understanding. In turn, these models can leverage this added information to better query user responses, summarize works, or generate ideas.
Depending on background and interest, work under this project can range from development of novel chemical descriptors, to training and testing of multi modal LLMs, to cheminformatics practices and structure prediction. Work will be fully remote. A background in chemistry, physics, computer science, or any other science/engineering is preferred. Linux/Unix experience and programing experience in Python or C/C++ is preferred but not required.
