Products
Foundations you license and build on.
4 modules, each a clean header-only C++17 library with an Eigen-only baseline and a self-checking demo. You compile it natively, embed it in your product, and ship that product to your own customers.
01 / Quant
Pricing, risk, and portfolios.
A deep derivatives and risk engine for desks, funds, and treasury tools. Build trading systems, risk dashboards, or analytics products on a foundation that already proves its own numbers.
Options
Heston, SABR, SVI, American, and exotic pricers, with copulas and tail dependence.
Risk
Extreme-value theory (stationary and non-stationary), stochastic volatility, VaR / CVaR.
Portfolio
Markowitz, tangency / max-Sharpe, risk parity (ERC), Black-Litterman, efficient frontier.
Backtest and rates
Kupiec and Christoffersen VaR backtests, cointegration, fixed income, credit and rates derivatives.
02 / Genomics
Population structure at scale.
A matrix-native pop-genetics foundation with a packed genotype store. Build research pipelines, breeding tools, or clinical-cohort QC on data that loads in seconds.
PCA and GWAS
GRM / SVD pop-structure PCA, OLS and logistic-IRLS GWAS with PC covariates and genomic-control lambda.
Selection scans
Per-variant and windowed Fst, Tajima's D, Fis, Watterson theta, HWE exact, LD (r-squared and D-prime).
Genotype store
Compact int8 and 2-bit packed dosage matrix, lossless, computing on the compressed store.
Readers
VCF, bgzipped .vcf.gz, PLINK .bed/.bim/.fam, and FASTQ quality control.
03 / Climate and environment
Trends, extremes, and attribution.
A spatial-climate foundation that reads NetCDF directly. The same engine powers agriculture (seasons, yield), water (floods, droughts), and energy.
Spatial analysis
Area-weighted EOF and teleconnections, with a zero-dependency NetCDF-3 reader (optional NetCDF-4).
Detection and attribution
Optimal fingerprinting (total-least-squares) with scaling factors and confidence intervals.
Trends and extremes
Mann-Kendall and Sen's slope, non-stationary extreme-value return levels, SARIMAX.
Interpolation
Empirical variogram and ordinary kriging for gridding sparse station data.
04 / Neural / AI
Self-hosted deep learning.
A gradient-checked tensor-autograd engine: transformers, LSTM and CNN, deep reinforcement learning, an anomaly and fraud stack, and a self-hosted LLM. It has its own page.
How you license it
OEM and embed, not naked resale.
License the foundation, compile it into your own product, and distribute that product, including as a hosted or SaaS service. End-users get your product, not the source. Like SQLite, Qt, or Unreal. You may not resell the source as a competing SDK.
See pricingAlso available on request
Standard ML, Finance, Bio.
Beyond the four headline modules we maintain a standard ML library, a finance and insurance module (short-rate models, Black-Scholes, QMC), and a neuro / behavioral TD-learning module. Ask us if your work needs them.
Ask about a moduleReady to build on it?
Tell us your module and your stack. We run a live demo, show the SHA-256 of the exact build, and deliver on payment.