Open source · MIT / Apache-2.0

AI infrastructure
built to be understood

Physical AI in Rust. Explainable AI in your data pipeline. Transparent models across every modality.

GitHub
11
Rust crates
6 FPS
on CPU
0
GIL
MIT
Apache-2.0
5
open-source projects

Projects

Five tools, one mission

Transparent AI from sensor to decision — open source, production grade.

ndarrow

The foundation: a zero-copy bridge between Apache Arrow and ndarray. O(1) conversions — views and ownership moves, never a data copy.

RustApache ArrowZero-copy

ndatafusion

nabled-powered linear algebra and ML as DataFusion SQL — 147 scalar UDFs for SVD, QR, eigen, PCA, and matrix functions, callable straight from a query.

RustDataFusionSQL UDFs

nabled

11 crates covering linear algebra, ML, robot kinematics & dynamics, control theory, sensor fusion, and full simulation pipelines.

RustPhysical AIRobotics

napparent

The Barn Effect algorithm in Rust. Appends {col}_effect interpretability columns to any Arrow RecordBatch — zero-copy, no GIL.

RustApache ArrowBarn Effect

Psychic Barnacle

The original MontOps explainability platform. SLiCKG architecture across text, tabular, image, video, and documents. 6 FPS on CPU.

PythonMulti-modalRAG-ready

Why MontOps

We build the tools we wish existed

Rust-first, production-grade

nabled and napparent have no GIL, no runtime overhead, no approximations — just fast, safe Rust built on ndarray and Arrow.

Explainability without dataset limits

The Barn Effect streams data in chunks. SHAP and LIME can't scale past a few thousand rows. We have no ceiling.

Arrow-native data pipeline

napparent operates directly on RecordBatches. Plug into DataFusion, Polars, or any Arrow-native stack with zero copies.

Full Physical AI stack

From linalg to simulation — nabled covers kinematics, dynamics, control, and sensor fusion in one coherent Rust ecosystem.

Transparency by architecture

The Barn Effect and SLiCKG aren't interpretability wrappers. Every prediction is auditable from first principles.

Open source, enterprise ready

MIT/Apache-2.0 across all projects. Enterprise consulting, custom deployment, and priority support available.

Building something that needs transparent AI?

Whether it's a robotics stack, a data pipeline, or a multi-modal ML system — let's talk.

nik@mont-ops.com