Quant Finance · Collaborator · Claude Opus
Generate, test, and validate alpha signals — at machine scale.
Three-agent architecture for systematic alpha generation. Idea Agent proposes market hypotheses from financial theory and alternative data signals. Factor Agent constructs formulaic alpha expressions with regularisation (originality enforcement via AST similarity, hypothesis-factor alignment, complexity control). Eval Agent backtests with walk-forward validation and provides feedback. The loop iterates until signals pass: IC > 0.02, turnover constraints met, decay profile acceptable, and correlation with existing alphas below threshold. Handles traditional factors (value, momentum, quality) and alternative data (web traffic, job postings, patent filings, satellite imagery, credit card transactions).
hypothesis-generatoralpha-constructorast-regularizerwalk-forward-backtesterfactor-evaluatoralternative-data-parsersignal-decay-monitorperformance-attributor| Bloomberg Terminal API |
| SEC EDGAR |
| Polygon.io |
| Alpha Vantage |
| SimilarWeb API |
| Alternative data feeds |
| Google Sheets |
| Memory |