Research
The organisation conducts applied research on artificial intelligence and large datasets. The goal is new knowledge and open methods, not commercial gain.
Project “Mukha”
A research project testing whether machine-learning methods can find persistent patterns in the microstructure of crypto-asset markets: the order book, trade flow, funding rates and liquidations.
Fly Trader: how it works
An animated demonstration: a model of the fruit-fly brain “reads the market news” and, when the terminal signals, ranks coins from strong to weak. The page shows neurons firing, energy per decision compared with a language model, research progress and a scientific description with links to the authors of the idea, the FlyWire data and the papers.
About the project, plainly
- This is research. We promise no returns and raise no funds.
- Hypotheses are registered before testing; results are published as they are, including those that were not confirmed.
- The project is not investment advice.
Partner and data
Project partner: Kavern Invest LLC (EDRPOU 44471062). The partner gives the organisation free access to its Prizma platform (prizma-orderbook.com) for the Mukha research project. [to be confirmed: agreement or memorandum, date]
Research data: public market data from the Binance and Bybit exchanges — order book, prices, funding rates, liquidations and trade flow.
Research directions
- Machine learning on time series: detecting signals in supply and demand imbalance.
- Computer vision: recognising structures on order-book heatmaps that human analysts read by eye. [to be confirmed: whether a prototype exists]
- Rigorous measurement: control samples, blind-feature checks, protection against overfitting to history.
What we need
Computing power for model training (GPU, cloud resources). We are looking for grant programmes that provide compute resources for research by non-profit organisations. [to be confirmed: GPU hours and memory required]
For cooperation: [email protected].