Bernie Boscoe

About

The Boscoe Research Group is based at Southern Oregon University and collaborates with students, research software developers, and domain scientists at SOU and other institutions.

Our work focuses on computational infrastructure for scientific collaboration: the systems, software, and workflows that allow research groups to work with data and knowledge over time. Current projects span scientific knowledge systems, retrieval-augmented language models, computer vision, ecological monitoring, and reproducible research software.

Two projects anchor the group’s current research: AquiLLM, an open-source system for preserving and accessing formal and tacit knowledge in research groups, and Green Crossing AI, an interdisciplinary computer vision and research software effort supporting long-term wildlife monitoring in Southern Oregon.

Research

Two research areas, one common problem: building computational infrastructure that remains useful inside real scientific collaborations.

Scientific knowledge systems

AquiLLM

AquiLLM is an open-source, modular RAG-LLM system designed to help research groups capture, preserve, search, and use both formal and tacit knowledge.

The project investigates retrieval, multimodal scientific collections, privacy-aware local deployment, open-weight models, research-group memory, and long-term knowledge continuity. AquiLLM is collaboratively led by Bernie Boscoe at SOU and Tuan Do at UCLA.

Funding: NSF CRII Award #2448094 · Alfred P. Sloan Foundation

Project website → GitHub →

Computer vision + ecology

Green Crossing AI

Green Crossing AI develops computer vision methods and research software for long-term wildlife monitoring, bringing together computer science and environmental science at Southern Oregon University.

The project uses camera-trap imagery, reproducible data pipelines, and computer vision to support research on wildlife movement and ecological change, while giving undergraduate researchers substantive roles in scientific software and data analysis.

Funding: WILDLABS + Arm

Project website → GitHub →

People

The Boscoe Research Group includes SOU students and researchers working within broader interdisciplinary and cross-institutional collaborations.

Funding & Support

Research in the group has been supported through competitive research funding and national cyberinfrastructure programs.

National Science Foundation

Alfred P. Sloan Foundation

WILDLABSWILDLABS + Arm

Selected Publications

Selected work spanning research software, scientific AI, astronomy, and environmental science.

AquiLLM: a RAG Tool for Capturing Tacit Knowledge in Research Groups

Chandler Campbell, Bernie Boscoe, Tuan Do.

US-RSE 2025 · arXiv:2508.05648

GreenCrossingAI: A Camera Trap/Computer Vision Pipeline for Environmental Science Research Groups

Bernie Boscoe, Shawn Johnson, Andrea Osbon, Chandler Campbell, Karen Mager.

PEARC 2025 · Project page

Learning the Evolution of Physical Structure of Galaxies via Diffusion Models

Andrew Lizarraga, E. Jiang, Jacob Nowack, Y. Li, Y.N. Wu, Bernie Boscoe, Tuan Do.

NeurIPS Machine Learning for the Physical Sciences Workshop, 2024 · arXiv

Photometric Redshifts for Cosmology: Improving Accuracy and Uncertainty Estimates Using Bayesian Neural Networks

Evan Jones, Tuan Do, Bernie Boscoe, Jack Singal, Yujie Wan, Zooey Nguyen.

The Astrophysical Journal, 2024 · arXiv

Selected Talks & Posters

• AI education and research with national infrastructure: lessons from AquiLLM and NAIRR. CIAO Cyberinfrastructure Alliance for Oregon Workshop, Portland, 2026.
• AquiLLM: a RAG-LLM for Capturing Tacit Knowledge in Research Groups. CounterBalance — Institutions and Interpretive Authority, Santa Fe Institute, 2026.
• Teaching Scalable Wildlife Image Processing with NAIRR Jetstream2 GPUs to Undergraduates. NAIRR Conference, Arlington, 2026.
• Green crossings: Student/Machine Learning in the Life Sciences. 4S Conference, Amsterdam, 2024.
• Arrays from the Sky: Astronomy and Data Science/Computer Science Collaborations. Willamette University, 2024.
• Scaling Up: Incorporating HPC Experience into Undergraduate Data Science using Gateways. ADMI, Atlanta, 2024.

Contact