Phorecast AI
A SaaS platform that generates 20+ year financial projections for early-stage drug candidates. Built for biotech founders and life science investors.
Visit phorecastai.com →The Problem
Biotech investors make multi-million dollar decisions based on Excel models held together with tape and prayer. A single drug candidate has hundreds of variables — epidemiology data, clinical trial probabilities, market dynamics, funding round dilution — and they all interact.
The client had a working Excel model, but it was fragile: circular references broke regularly, assumptions were buried in cell formulas, and sharing scenarios meant emailing spreadsheet versions back and forth.
What I Built
Revenue Forecasting Engine
27 calculation modules chained into a single pipeline. Epidemiology → market sizing → revenue projection → financial modeling. Each module is independently testable.
MOIC & Cap Table Solver
Iterative solver for equity dilution across funding rounds with circular dependencies. Multiple On Invested Capital calculations that match the client's Excel model to the decimal.
Professional PDF Reports
Auto-generated investment reports with KPI chart grids, custom formatting, and client branding. The kind of output investors expect to see in a board meeting.
Transparent Data Sources
846-row parameter metadata system. Every number traces back to its source — UN population projections, NHANES epidemiology, drug class benchmarks. No black boxes.
The Hard Part
The circular reference solver.
In pharma financial modeling, equity dilution depends on valuation, which depends on projected revenue, which depends on funding allocation, which depends on equity dilution. It's circular by nature.
Excel handles this with iterative calculation (and occasionally crashes). I built a convergence solver that resolves these dependencies reliably, hitting 100% parity with the reference model — and in the process, found a formula error in the client's original Excel that had been there for months.
Stack
Status
Live and in beta with early users since February 2026. The platform handles multiple therapy areas and drug classes, with the calculation engine running entirely client-side for speed.