About This Work
Both projects deal with information — reading it, finding what matters, making it useful. One analyses documents at scale. The other connects AI to everything I'm working on day to day. The consistent interest is what actually holds up in practice, not what looks good in a demo.
Built for daily use. Running continuously.
A personal AI system that connects to my notes, calendar, and health data. Background processes run continuously to surface patterns and flag things worth attention. Conversational interface that can read and write across all of it.
Reads all my markdown files, maintains a full-text search index, scores and ranks what's worth attention.
Conversational AI agent with tool use — reads and writes notes, queries calendar, health data, and long-term memory.
Runs continuously: extracts a chronicle of significant events, synthesises patterns, surfaces priorities, verifies claims.
Google Calendar, health data via Google Drive sync, Gmail with draft approval workflow.
Private project. Not currently public.
Built to understand how you extract structured intelligence from unstructured documents at scale. Four-tier analysis pipeline — free signal extraction through LLM synthesis — with a critique layer to catch hallucination propagation. Real working code, AGPL licensed. Domain-agnostic by design: the same pipeline works for legal discovery, research synthesis, or operational incident analysis.
Zero-cost preprocessing. Structural patterns, word frequencies, temporal references, entity extraction.
LLM-powered analysis. Discrete insight extraction from individual documents.
Correlation analysis. Recurring themes, contradictions, temporal evolution across documents.
Deep synthesis. Reports, recommendations, actionable intelligence.
Python, provider-agnostic LLM orchestration, SQLite indexing. AGPL-licensed core with domain-specific adapters.
⟶ Natural language interface for document intelligence
⟶ Transform from analysis tool to intelligence platform
⟶ Professional interface ready for production
⟶ Major reliability breakthrough