Context

A quant research team at a leading systematic fund was using SEC filing data in their trading models. Their issues were simple but painful and persistent. Very often, the data would be missing, inaccurate, or have typos.

Issues

The team's existing data vendor had proven unreliable, with frequent errors requiring manual verification and multi-day correction cycles through phone tag. Without better options, the QR team personally hunted down data issues for 3+ hours every morning. Their data ingestion process would implement basic outlier detection and would send an email detailing failed checks to the QR team every morning. Every single morning, the QRs needed to refer to this email, compare the flagged data points with publicly available references, and play phone tag with the data vendor to get them to reissue the feed with corrections. All before they could get to their day job of quant research.

sieve solution

We built an API endpoint that lets the team query for specific data from financial filings. The requesters only need to provide the relevant ticker, reporting period, and desired metrics. Behind the scenes, sieve finds the right document, uses AI to extract the requested data points, and forwards the data to a team of human reviewers. After careful human review to ensure accuracy, the data is returned to the requester via the API. This replaces hours of researcher time hunting down and verifying data every morning, and lets them focus on their research.