The market-data pipeline was rebuilt and broadened beyond raw price history to include liquidity, macro and on-chain sources. On top of that stronger foundation, the first baseline ML training cycle was completed on a reference market — an early test of whether the data pipeline could actually support a full research cycle.