Technical Intelligence Corpus

Knowledge Graph-guided Retrieval-Augmented Generation (KG2RAG): uses a knowledge graph to improve retrieval for RAG, positioning entity/graph signals to steer context selection and generation.

Automated recon pipeline for a target domain that orchestrates multiple scanning tools to enumerate assets and identify potential vulnerabilities.

Provide an MCP (Model Context Protocol) server that exposes Todoist capabilities by implementing the Todoist REST API and supporting the Todoist Sync API.

Provide a benchmark (and likely associated evaluation methodology) for measuring how effectively multimodal large vision-language models (LVLMs) perform “unlearning” of copyrighted visual content (e.g., characters/logos) after training, addressing weaknesses in existing evaluation approaches for multimodal cross-modal memorization and removal.

Task-conditioned uncertainty-aware costmaps for legged locomotion by predicting foothold/terrain feasibility while explicitly modeling epistemic uncertainty conditioned on terrain observations and commanded task/context, then using that uncertainty to shape costmaps for motion planning and path selection.