Collimator is a Python‑based simulation and modeling platform that unifies physics‑based modeling, automatic differentiation, and machine‑learning tools for complex dynamical systems. It enables engineers to build, simulate, and optimize models—including surrogate models, digital twins, and model‑predictive control—using JAX‑accelerated compute, GPU support, and integrated code generation for deployment.
Funding
$5M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Product
Problem
Engineers developing complex dynamical systems often rely on separate tools for modeling, simulation, optimization, and machine‑learning integration, leading to fragmented workflows, duplicated effort, and difficulty scaling analyses across high‑performance compute resources.
Solution
Collimator provides a unified, Python‑based platform that combines physics‑based modeling, automatic differentiation, and machine‑learning pipelines to design, simulate, and optimize dynamic systems in a single environment. Built on JAX, the suite supports surrogate modeling, digital twins, system identification (e.g., SINDy), model‑predictive control, and physics‑informed learning, all accessible through a collaborative cloud or local interface. Users can run large‑scale parameter sweeps, Monte Carlo simulations, and GPU‑accelerated computations, while the integrated optimizer leverages auto‑diff for auto‑tuning, local and stochastic optimization strategies. The platform also includes tools for code generation, data visualization, and an AI chat assistant that can generate code, explain models, or create models from equations, streamlining the development cycle from concept to deployment.
Target Audience
Primary users are control engineers, system designers, and researchers in autonomous vehicles, robotics, signal processing, and wireless communications who need an end‑to‑end environment for modeling, simulation, and optimization of complex dynamical systems.
Features
- General‑purpose model editor with Python and JAX integration for high‑performance, differentiable simulations
- Automatic differentiation‑driven optimization supporting auto‑tuning, local, and stochastic methods
- System identification tools (e.g., SINDy) for data‑driven plant model discovery
- Machine‑learning modules for physics‑informed learning, digital twins, and neural network training with flexible data pipelines
- High‑performance compute support including parallel distributed simulations, GPU acceleration, and large‑scale parameter sweeps
- C code generator and export utilities for deploying models to embedded or real‑time systems
- Integrated data visualizer and AI chat assistant for model inspection, code generation, and interactive assistance