Lumina Engine

A single-point DCF hides how sensitive a valuation is to its inputs. Small changes in growth or discount rate move the answer a lot, and bad source data can silently produce a confident wrong number.

• —Pulls a company's historical financials from SEC EDGAR, with Alpha Vantage and FRED for market and macro inputs.
• —Projects revenue with a cohort-decay engine anchored to the company's own growth history.
• —Discounts cash flows at a CAPM-derived cost of capital.
• —Runs 5,000 Monte Carlo trials to produce a valuation distribution rather than a single figure.
• —Reverse-DCF solves for the perpetual growth rate implied by the current market price.

• —Validation runs before any calculation, so data inconsistencies are caught and flagged instead of flowing into the valuation.
• —Edge cases are surfaced to the user rather than smoothed over, to prevent silent valuation errors.
• —A FastAPI backend keeps the numerical work in Python; the Next.js front end only presents results.

• —23 passing tests, run in CI on every push.
• —MIT licence.

Shipped and still being iterated on. The repository goes public shortly after this site launches.