Area 02
Finance & Economics
Rigorous inference under regime change — modeling the non-stationary systems most market work quietly assumes away.
What this area is
Markets do not sit still. The hard problem behind most market behavior is non-stationarity — regimes that shift underneath any model trained on the last one. We work that modeling problem first and let applications follow, rather than the other way around.
A second thread treats market data itself as a kind of language: tokenized sequences a model can be trained to read. That is also where this area meets the lab's model-architecture research — markets are a demanding, honest substrate for testing what a sequence model can really learn.
Lines of work
Markets Research
Non-stationary stochastic time-series modeling. Techniques for inference under regime change — the modeling problem most market work assumes away. Modeling first; instrument-class application is downstream of the methodology.
Read moreMarket Glyphs
Treating market behavior as a language: tokenizing price and flow into sequences a model can read, and testing where sequence-model architecture earns its keep on real, non-stationary data. The applied edge of the lab's model-architecture work.
Detail page forthcoming