About the product
Price the book. See the risk. Own the stack.
Numeraire is an application for pricing equity options and managing counterparty credit risk (CCR) end to end — from overnight mark-to-model through greeks and PnL to expected exposure and PFE. A C++ engine does the heavy lifting; a Django journal lets you review the book without opening a notebook.
What Numeraire is
Numeraire is a derivatives pricing and risk platform built around a C++ engine. The engine prices instruments, computes greeks, builds discount curves and vol surfaces, and writes official marks plus EE / PFE exposure into SQLite. A Django and Bootstrap journal sits on top so you can review the overnight book, challenge a number, and see what the library can already price.
Today the production focus is equity options and CCR. Commodity futures are under active development — listed outrights and session settles are coming online; options on futures and a fuller commodity stack are next.
What-if
On a live trade you can bump a single risk factor — spot, implied vol, rate, dividend yield or time to expiry — and reprice without touching the official mark. Shocked PV and greeks sit next to the EOD baseline so you see what that move does to the position, and how it would ripple through the book.
Same market inputs the overnight engine used; nothing is written back. Reset and the mark of record is still the mark of record.
Trade view + market data
Open a trade and see the mark next to the market that priced it — underlier spot, discount curve and implied volatility for options. Equity and options feeds land automatically from Massive (Polygon); after prep, official MTM is recomputed so you stay on analysis, not plumbing.
Surfaces and term structures live beside the book: challenge a greek or PV against the same curve and vol input the engine used.
CCR exposure
After MTM, Numeraire simulates risk-factor paths and persists expected exposure and potential future exposure (PFE 95% / 97.5%) for the book. The journal surfaces those profiles so counterparty risk sits next to the marks — not in a separate spreadsheet.
Stack
C++20 pricing core, Python and Django for the web layer, SQLite storage, ECharts for curves, surfaces and exposure. Hosted on AWS in production. Quant Lab is open without login so anyone can try the pricing ideas; the live book, overnight MTM and market views stay behind sign-in.
Development
Two servers, one repo. Dev is where the journal gets
broken on purpose. Prod is the live book — same code,
after CI/CD. A merge to main ships itself: GitHub Actions
SSHs in, rebuilds the C++ engine, syncs Python, migrates, restarts.
The application is developed in that spirit — not a laptop demo that
later “gets deployed”.
Local SQLite
The book lives in a local SQL file — not a warehouse in the cloud. Production trades, Massive market data, official MTM and EE / PFE exposure are all written there: the same store the C++ engine persists and the Django journal reads.
One file, one source of truth. Open it, query a mark, and see the trade next to the market that priced it.
Unit & integration tests
The C++ engine and the journal have solid unit and integration coverage — pricers, bootstraps, MTM persistence, exposure. That is how the methodology stays honest: a wrong formula or a silent regression is caught before it hits the book.
Disclaimer
Numeraire Journal is provided for demonstration, education and evaluation only. It is not an offering of investment advice, a licensed trading system, or a product intended for commercial production use. No warranty of fitness for any particular purpose is made; use is at your own risk. For access to the full journal, contact the owner.