Technical OSINT Corpus

Agentic RAG to support file-level bug localization in software maintenance pipelines (aid root cause analysis/triage and improve downstream stages like statement-level localization and patch generation).

An MCP server/tool that scans remote MCP documentation for prompt-injection vulnerabilities and supports LLM security testing.

Research code/paper implementation studying “state contamination” in memory-augmented LLM agents, specifically the failure mode where toxic/adversarial context can be “memory laundered” via summaries that evade standard detectors while retaining harmful framing for later agent behavior.

Web app that transcribes user-uploaded speech audio into text using OpenAI Whisper (multilingual + auto language detection) and offers text-to-speech via gTTS, wrapped in a Gradio UI for near real-time use.

Neural decoding pipeline that attempts to reconstruct dream narratives from EEG sleep data.