Quant / 01

Quant Infrastructure

The data and compute behind the factor research. DC17 schedules work and stores multi-source quantitative data from 1925 to 2026; ten cloud nodes form a standing compute pool that runs mining and backtests alongside a university HPC cluster and on-demand GPUs.

Standing cloud vCPU · 10 nodes: 160
Methods & tools: ClearML · Slurm · Apptainer · Parquet · CUDA

Research visuals

Compute topology and node scheduling · DC17 holds the experiment plan and schedules it through ClearML; seed tasks enter a shared queue, the ten cloud nodes pull them on demand, and results return to DC17. The university cluster is requested by queue, and GPUs accelerate core operators with CUDA. The animation illustrates the flow, not real timings. — **Compute topology and node scheduling**
DC17 holds the experiment plan and schedules it through ClearML; seed tasks enter a shared queue, the ten cloud nodes pull them on demand, and results return to DC17. The university cluster is requested by queue, and GPUs accelerate core operators with CUDA. The animation illustrates the flow, not real timings.
View full image ↗

Quant data foundation · Multi-source data is aligned on permanent security IDs and disclosure times, so each backtest slice only uses information public at the time. Row counts reflect the archive as of November 2026. — **Quant data foundation**
Multi-source data is aligned on permanent security IDs and disclosure times, so each backtest slice only uses information public at the time. Row counts reflect the archive as of November 2026.
View full image ↗