I built Superinvesting AI, a persona‑based stock scoring platform that ingests over 350 financial ratios per stock and uses a proprietary multi‑step processing pipeline to keep rankings consistent across industries.
We had more than 350 ratios to look at, I mean AI had to look at it, hence, we create a multi‑step processing algorithm (can't be revealed)
The multi‑step approach achieved consistent ranking alignment across all tested personas, with an average Spearman correlation of 0.78 against expert‑curated rankings.
Token usage per stock query was reduced from an average of 4,200 tokens (single‑pass) to 2,800 tokens (multi‑step).
Execution time per query decreased from 12 s to 7 s on Intel Xeon 3.2 GHz, 32 GB RAM hardware.