Digital Asset Credit Score: Wallet-Level Risk Modeling

Overview

This was my main project in Aetherum.ai, where I built a credit scoring algorithm to assess user credit-worthiness alongside feedback from Professors at the Haas School of Business. The financial model aggregates conventional credit scoring metrics: The FICO Score, The user's AUM and the user's liquidity adjusted net-worth. Alongside the conventional credit scoring metrics, I also implemented 2 metrics: the volatility score and the asset quality which mainly evaluates the user's digital asset portfolio in terms of the implied volatility and the quality of the digital asset itself. The final score is a weighted sum of the conventional credit scoring metrics and the 2 additional metrics. The score is then fed into the Aetherum's agentic workflow for the final loan underwriting process. I focused on making the system practical for real-world decisioning: clear feature definitions, transparent scoring logic, and evaluation workflows to tune thresholds and validate stability under changing market conditions. The result is a score designed to be interpretable, testable, and usable as an input to collateral parameters and portfolio risk monitoring.

Highlights

• Designed a wallet-level scoring framework (DACS) combining on-chain behavior features with market risk signals
• Built a Python feature pipeline for cleaning, transforming, and aggregating wallet metrics into model-ready inputs
• Created evaluation + backtesting workflows to tune score thresholds and sanity-check stability across market regimes
• Produced interpretable outputs to support underwriting decisions and risk parameter setting for collateralization