Heron Finance
Fintech
Lead Product
Heron Finance
ML-powered robo-advisory for Private Credit, B2C and B2B.
About the Project
When I joined Heron Finance, the company had a bold vision: make private credit as accessible and intelligent as public markets. But the challenge was clear — investors were flying blind. Portfolio construction was manual, risk data fragmented, and credit performance opaque.
My role was to turn that problem into a product — designing the first machine-learning-powered platform that could price, score, and diversify loans autonomously.
The Challenge
Private credit was booming post-2020, yet its infrastructure looked decades old. Most investors and advisors were still relying on static PDFs, intuition, and inconsistent fund data. We needed to solve two core problems:
How can we lower the barrier to entry for everyday investors in private credit?
How can we build trust that a robo-advisor could analyze fund data with the rigor of a human financial planner?
The solution had to bridge credibility and automation — blending intelligence with intuition.
My Approach
I led product discovery alongside founders, engineers, credit team, and early design partners to reverse-engineer how institutional credit teams made decisions. We interviewed fund managers at Ares and Golub, analyzed 50+ fund portfolios, and identified the signals that drove confidence: performance history, sector correlation, and loan-level volatility.
From that insight, I defined the 0→1 roadmap:
Phase 1: Build an ML-driven risk engine that parses and prices private credit funds automatically.
Phase 2: Layer on investor experiences — automated portfolio generation, personalized dashboards, and dynamic yield forecasting.
Phase 3: Establish feedback loops to continuously tune accuracy and user trust.
Execution & Collaboration
I partnered closely with engineering and design to create the Synthetic Loans model — a dynamic data pipeline that transformed raw fund statements into loan-level exposure and automated re-valuation.
Together, we:
Designed a valuation model that simulated portfolio diversification and rebalanced in real time.
Re-architected onboarding to make complex credit data feel human-readable and intuitive.
Established bi-weekly “ML Reviews” where engineers, PMs, and data scientists aligned on prediction accuracy and hypothesis testing.
It was a rare intersection of finance, product design, and machine learning — all built within a startup’s speed.
Outcome & Impact
We launched Heron Finance in November 2024, scaling to over 1,200 loans and driving a 10× growth in AUM in the first month post-launch.
The ML model improved default prediction accuracy and reduced loss exposure by ~$100K per month.
User activation grew 163% after redesigning onboarding and portfolio visualization.
And beyond metrics, Heron’s system created an entirely new category of intelligent private-credit investing — one where individuals could access institutional-grade diversification in seconds.
Lessons Learned
Building Heron was a crash course in aligning humans and algorithms.
I learned that trust in automation isn’t built through simplicity — it’s built through clarity. The more we explained the intelligence behind the model, the more users believed in its outcomes.
We didn’t just build a risk engine; we built confidence in a new way to invest.
