Foundations, not slideware
C++ computation foundations for your industry.
Built for developers, ready for anything. Licensed, self-hosted modules you compile, embed, and ship inside your own products, with the math already proven on real public data.
Proven, not promised
verified recovery checks pass
on real public data, SHA‑256 pinned
autograd gradient-check error
hidden dependencies, Eigen only
For engineering teams
Header-only C++17. One command to a running demo.
No framework, no Python runtime, no service to call. You get clean source you compile natively, the same class of engine behind high-frequency trading and genomics research, tuned to run fast on ordinary hardware.
- Eigen-only baseline, no hidden dependencies
- AVX2 floor, auto AVX-512 where present
- Cache-blocking, mmap, OpenMP, LTO
- Every module ships a self-checking demo
# compile a module's demo, native to your CPU
$ ./compile-native.sh quant_demo
# run it, it recovers a known answer
$ ./quant_demo
[PASS] Heston ATM call 10.4506
[PASS] VaR backtest (Kupiec) ~1% exceptions
[PASS] risk parity (ERC) equal to 3e-16
What we provide
C++ computation modules your team builds on.
License a foundation and ship your own product on top, like SQLite, Qt, or Unreal. Need it shaped to your stack? We tune it to your exact CPU, write custom modules, and take bespoke missions.
Core features
Everything you need to build faster.
Performance layer
SIMD (AVX2 and auto AVX-512), cache-blocking auto-tuned to your CPU, mmap RAM-resident reads, OpenMP and LTO. The speed is part of the product.
Verified math
Every module recovers a known answer. 50 / 50 recovery checks pass, built NaN-honest so a silent bug fails its assert instead of hiding.
Real-data proof
Validated on real public files a domain expert recognizes, every input pinned by SHA-256 and source URL, so anyone can re-download and re-check.
Self-hosted
100% C++17, Eigen-only baseline, zero Python. It runs entirely on your machines, nothing leaves your network.
OEM / embed license
License the foundation, compile it into your product, and ship that product to your own customers, including as a hosted service.
Clean delivery
You receive only your licensed module, a self-checking demo, and a portable build script. Watermarked to you, SHA-256 verified at handover.
Use cases
Foundations you build on, across industries.
A module is rarely single-industry. The same climate engine powers agriculture and water; the same anomaly stack powers fraud, fault detection, and quality control.
Quant
Pricing, risk, and portfolio engines
Options (Heston, SABR, SVI, American, exotic), copulas, extreme-value risk, VaR backtesting, portfolio optimization. Build trading tools, risk dashboards, or treasury analytics on top.
Genomics
Population structure at biobank scale
PCA, GWAS, Fst selection scans, a packed genotype store reading VCF, PLINK, and FASTQ. Build research pipelines, breeding tools, or clinical-cohort QC.
Climate and environment
Trends, extremes, and attribution
EOF and teleconnections, optimal-fingerprint detection and attribution, non-stationary extremes, kriging. Powers agriculture (seasons, yield), water (floods, droughts), and energy.
Neural / AI
Self-hosted deep learning and anomaly
A gradient-checked tensor-autograd engine: transformers, LSTM, CNN, deep RL, a self-hosted LLM, and an anomaly and fraud stack. See the AI suite
Who builds this
Two freelancers, not a company.
We are Co Bros and Alex Ramanaidou, two independents who believe rigorous mathematics is the only honest way to answer hard data questions. We build the foundation and prove it on real data so our clients can compare fast. Open to custom work.