Long Running Language Model Thread Truncation
U.S. Publication 2025/0111847Manages LLM conversation context at scale by clustering prompts and responses by topic and truncating stale clusters via reference scoring.
I believe creating patents are practice for thinking. They train you to look at a system, find the seam where it isn't working, and articulate what would close it.
A small sample of named filings from the portfolio — the ones I'd point at first if you asked what I actually work on.
Manages LLM conversation context at scale by clustering prompts and responses by topic and truncating stale clusters via reference scoring.
Generative adversarial network dynamically creates personalized UI elements based on inferred user interests at runtime.
NLP and sentiment analysis across enterprise conversations to detect product dissatisfaction and recommend alternatives.
ML-driven agentic automation that recognizes task patterns, checkpoints for human approval, then completes offloaded work.
The portfolio spans twelve technology domains. Click any row to read the through-line and see every patent in that domain.