tegridydev
opensource dev | independent researcher | nerdstuff
I build local tools, open datasets and practical research projects around security, language models and things that interest me.
Most of my work starts as a question or a tool I wanted for myself. If it turns out useful, I clean it up and put it online.
That’s pretty much it.
interests AI security · Local LLMs · Dataset engineering · Mechanistic interpretability · Multi agent systems · Cool stuff · OSINT · Physical security systems
Selected work
- auto-md ↗ Turn files, folders and GitHub repositories into clean Markdown for LLM and RAG workflows.
- abstract-agent ↗ Local multi agent research workflow for exploring literature, novelty and hypotheses.
- Mechanistic Interpretability ↗ Research notes and open tools for inspecting transformer internals. MechaMap ↗
- multi-agent-secops-llm ↗ Experiments with LLM agents for security operations and analysis.
- physical-security-dev ↗ Practical notes for CCTV, access control and physical security development.
- Python OSINT Notebook ↗ OSINT tooling, Python automation and intelligence workflows.
- dnd-llm-game ↗ Local multi LLM role playing and interactive storytelling.
Research and data
- Research Papers ↗ Academic documents for OCR, classification and multimodal research. DOI ↗
- Open-MalSec ↗ Open cybersecurity data covering phishing, malware, vulnerabilities, fraud and hostile communications. DOI ↗
- Infosec Tool Output ↗ Structured security tool output for blue team and security agent research. DOI ↗
Writing
Start here
- Reading LLM Model Maps: Activations and Causal EvidenceRead a model map without confusing architecture, measured activations and causal evidence, using a local viewer and explicit capture provenance.
- Dataset Cleaning Without Losing IDs or MeaningClean datasets with explicit row accounting, preserved identifiers, conflict quarantine and reproducible splits rather than silent type conversion.
- CloudVec: Vector Index Freshness with a SQLite OutboxFollow updates, deletion and crash recovery through a SQLite outbox, with freshness and replay checks kept separate from semantic search quality.
Latest writing
- Research Papers DatasetHow and why I built the research papers dataset.
- Xanadu Inspired Links to Versioned Document PassagesKeep document versions, passage spans and link identities intact through export and import, including Unicode text and historical source references.
- Organising Research Notes Without Losing HistoryCreate a useful reading copy of research notes while preserving original files, duplicate relationships, conflicting versions and attribution.
- Building Technical Knowledge Bases for TroubleshootingOrganise a technical knowledge base around the next useful check, with source evidence, device time and receipt time kept distinct.
- Embedding Compatibility: Models, Vectors and SearchRecord model revisions, pooling and normalisation beside embeddings, then compare retrieval methods without mistaking equal dimensions for compatibility.
Research citations
Feel free to reach out <3
Humans and authorised AI agents are welcome. Bulk outreach isn’t.
signal accepted