Vanessa Feng

Piped the probability‑of‑default model into the watchlist as a loan filter, saving about 10 hours of manual review a month.

• Tested 8 model thresholds against recall, precision, accuracy and ROC‑AUC; chose the one that kept recall above 0.90 with the other metrics at or above firm benchmarks.
• Built Datarails dashboards that segment about 300 flagged loans a month by product, geography, vintage and sponsor.
• Restructured the credit approval memo to quantify risk assessments; approval meetings ran about 50% faster.

Tailored model portfolios for six new clients to their risk profiles, liquidity needs and existing exposure, and tested their long-term goals with Monte Carlo simulation.

• Designed duration- and credit-appropriate fixed income strategies for 10+ client portfolio reviews with the financial planning team.
• Screened and modelled 10+ US large- and mid-cap stocks; built the Excel forecasts through 2028 that the team used for tactical positioning.
• Assessed 5 private market funds in infrastructure, real estate and private credit for suitability across 50+ client portfolios.

Designed the pitch deck that secured Rappo’s first round of funding.

• Built a database of 500+ target investors with Python and web scraping, with a risk-appetite scorecard for fit that improved targeting accuracy 50%+.
• Refined go-to-market from investor feedback; a new one-pager lifted the cold-email success rate 10%+ as Rappo won 100+ clients and $6,000+ in revenue in two months.

A logistic-regression model of credit card default. Vintage curves set a 6‑month performance window; roll-rate analysis tracked accounts within it.

• Narrowed the candidate variables with break-and-heal binning, weight of evidence and information value, stepwise regression and cluster analysis.
• Set warning and breach thresholds for ongoing monitoring.