Buildactive
LLMoxie
Large Language Model based Agentic Verification and Exploration of Natural Inquiry. Building trustworthy AI agents for scientific research workflows. (Formerly LLMaven.)
PythonAIAgentic SystemsResearch
Year2025RoleTeam memberOutcomePublished at the KDD 2026 SciSoc Agents & LLMs workshop
Overview
LLMoxie is an open-source AI control plane funded under NSF NAIRR Award #240292 (Schmidt Sciences VISS program). It provides AI infrastructure for scientific discovery across cloud and HPC environments.
Architecture
Three-layer design:
- Inference engine — Azure, AWS, GCP, or local vLLM
- API gateway — LiteLLM with authentication, rate limiting, and PII masking
- RSE-Plugins layer — Domain-specific agents and skills for scientific workflows
Related
Part of the broader SSEC Research Software Engineering research portfolio. Uses RSE-Plugins for domain-specific agent skills.