Developers

Install the SDK, call one function, get the 8 features that matter. Plug into scikit‑learn, pandas, and PyTorch.

pip install optimal8View Docs

Build Smarter AI with Less

The universal law of efficient intelligence. Build accurate models 100x faster, using 100x less compute—without sacrificing performance.

Get Started for Free Request Enterprise Demo
1
100+ Raw Features
→
2
Optimal‑8 Engine
→
3
8 Optimal Features
→
4
✅ Accurate, Lightweight Model

Why Optimal‑8?

⚡

Real‑Time Anywhere

Deploy on microcontrollers, phones, edge devices — negligible latency, no GPU required.

🧠

Domain‑Agnostic

Finance, healthcare, IoT, sports, security — one law, infinite applications.

🔒

Privacy‑First

Keep data on‑device. Optional cloud API when you want it.

📦

Ultra‑Lightweight

Models as small as 21KB — our Go AI runs on a mini‑PC.

Proven Across Industries

Game AI

21KB model playing Go on consumer hardware.

Financial Forecasting

High‑accuracy FX predictions with minimal features.

Healthcare

Diagnostics with optimized feature sets and lower compute.

Cybersecurity

Real‑time anomaly detection at the edge.

Ready to Build the Future?

Join developers and enterprises already leveraging Optimal‑8.

Start Building for Free

Quickstart

# 1) Install
pip install optimal8

# 2) Use
from optimal8 import select_features, O8Classifier

X8 = select_features(X, y)      # returns top 8 evolving features
clf = O8Classifier().fit(X8, y) # train a compact model

# 3) Predict
yhat = clf.predict(select_features(X_new))

Key Guarantees

  • Starts at exactly 8 core features
  • Maintains an evolving pool with weights
  • Prevents bloat; favors generalization
  • Transparent feature rankings & export
Note: API surface is illustrative — align with your actual SDK when published.

Cloud API

POST /v1/select
Authorization: Bearer <token>
{
  "X": [[...], [...]],
  "y": [...],
  "strategy": "o8-default"
}

# Response
{ "features": [3, 7, 11, 16, 21, 26, 32, 40], "scores": [...] }

Tokens, rate limits, and SLA depend on plan. Edge kits available for on‑prem.