Deterministic compression for structured signals. One geometric architecture, eleven applied domains.
Zer0pa builds codecs that compress domain-specific data — sensor telemetry, hand-pose streams, robot motion, geospatial trajectories, financial time-series, neural signals, biosignals, motion capture, speech prosody, digital ink, and multimodal content — using a unified set of eight geometric primitives.
Every codec guarantees bit-exact deterministic replay. No GPU required. No neural network. No probabilistic reconstruction. What you encode is exactly what you decode, every time.
| Package | Domain | Headline |
|---|---|---|
zpe-iot |
IoT / Sensors | 17× compression on real sensor datasets, edge-deployable (Rust + Python) |
zpe-xr |
XR / Hand Tracking | 56× compression, sub-mm pose fidelity, 0.025ms latency |
zpe-robotics |
Robotics / Motion | 187× compression on LeRobot data, wire-format replay |
zpe-geo |
Geospatial | Trajectory codec, sub-metre fidelity, zero dependencies |
zpe-ft |
Financial Data | OHLCV + tick-stream codec, ≥13× on synthetic benchmarks |
zpe-neuro |
Neural Signals | Extracellular spike-train codec, DANDI-validated |
zpe-bio |
Biosignals | ECG + EEG codec in Rust and Python, embedded reference path |
zpe-mocap |
Motion Capture | Joint-angle compression, search-without-decode |
zpe-prosody |
Speech Prosody | Pitch/rhythm/stress codec, 4/6 gates verified |
zpe-ink |
Digital Ink | .zpink stylus codec, >5× vs raw float32 |
zpe-imc |
Multimodal Platform | 10 modality lanes, 276.8k IMC words/s, integration layer |
All repos are source-available under SAL v6.0. Free below USD 100M annual gross revenue. Commercial, hosted, or legal-interpretation questions: architects@zer0pa.ai.
- Developers building on constrained devices, real-time pipelines, or audit-grade data infrastructure
- Investors and evaluators conducting technical due diligence on codec IP
- Grant reviewers assessing computational physics or information-theory research
- Deep tech engineers who want deterministic, verifiable compression — not black-box approximation
Pick the domain repo closest to your signal type. Each repo has a quickstart, proof artifacts, and benchmark data. ZPE-IMC ties the architecture together across domains.