RavenPack Launches NVIDIA Nemotron-Powered Pipeline for Rapid Financial Dataset Generation
RavenPack has introduced a new dataset generation pipeline on its Bigdata.com platform, designed to convert plain-language research hypotheses into backtest-ready datasets. By utilizing a fine-tuned 4B NVIDIA Nemotron model distilled from a larger 120B teacher model, the system can label massive volumes of financial text with high efficiency and accuracy. In performance testing, the pipeline labeled over 845,000 excerpts in under an hour using two NVIDIA H200 GPUs. This approach significantly reduces costs—labeling subsequent passes is 96% cheaper than using proprietary large-scale models—while maintaining high fidelity to the original teacher model's classifications. The technology allows hedge funds and asset managers to move from research ideas to scalable, signal-driven backtesting at a fraction of previous time and cost requirements.
Tickers: $NVDA
Topics: AI dataset generation Data Science Financial technology Machine Learning NVIDIA Nemotron