Flirting with Models What does a full-stack quant research platform and process look like?
18 snips
Feb 13, 2023 AI Snips
Chapters
Transcript
Episode notes
Explosion Of Structured Financial Data
- Data availability has exploded: historical fundamentals, pricing, options, TRACE bond data, and structured ownership datasets now exist.
- Meredith credits Moore's Law and cheaper storage for enabling wide, deep mosaics of structured inputs going back decades.
Value Judgement For Pointy Datasets
- Alternative pointy datasets like ESG, supply chain manifests, and executive/board data provide targeted signals.
- You must judge cost-benefit and vendor value-add (e.g., mapping manifests to companies) before integrating them.
Turning Unstructured Text Into Signals
- Unstructured data ingestion and NLP allow quantitative teams to ingest fundamentally written material at scale.
- Meredith describes scraping Moody's manuals to reconstruct financials and using Doc2Vec on 10-K MD&A to vectorize documents for grouping.