Positronic Robotics
About
About Positronic Robotics
Positronic Robotics provides an open-source Python stack that standardises how vision-language-action and other robot-learning models connect to real machines. A canonical task representation and codec layer translate between robot data and each model's expected format, allowing teams to change model vendors or hardware without rebuilding the collection and inference pipeline.
The workflow covers teleoperation in simulation or on hardware, dataset capture, task-specific fine-tuning, unified production inference, failure measurement and edge-case recollection. Positronic uses the same stack for PhAIL, a reproducible benchmark that tests models on identical Franka FR3 and Robotiq bin-picking hardware and publishes datasets, training scripts and evaluation methods. The toolkit supports OpenPI, NVIDIA GR00T, SmolVLA and ACT while keeping model and robot interfaces separate.
Team
- Sergey Arkhangelskiy - CEO and Co-Founder
Research
- PhAIL: A Real-Robot VLA Benchmark and Distributional Methodology
2026-05-28 - Research Publication - Launches a real-robot VLA benchmark on Franka FR3 with distributional metrics beyond binary success rates. Includes analysis code, paper sources, annotation pipeline, dataset metadata, and public dataset download instructions. Reports throughput-style comparisons across public VLA systems using time-to-success curves, bootstrap confidence intervals, and KS tests.
Source
Timeline
Events
- 2026-05-28 - Research Publication - PhAIL: A Real-Robot VLA Benchmark and Distributional Methodology
Launches a real-robot VLA benchmark on Franka FR3 with distributional metrics beyond binary success rates. Includes analysis code, paper sources, annotation pipeline, dataset metadata, and public dataset download instructions. Reports throughput-style comparisons across public VLA systems using time-to-success curves, bootstrap confidence intervals, and KS tests.
Source - 2026-03-31 - Product Launch - Positronic Robotics launches PhAIL benchmark to test real-world performance of physical AI systems
Positronic launched PhAIL as a real-hardware benchmark for foundation models in robotics, focused on throughput, reliability, and failure modes rather than simulation-only success rates. The company said leading models including OpenPI 0.5, GR00T, DreamZero, and SmolVLA were evaluated on physical robots doing commercial tasks, alongside human and teleoperated baselines. Positronic framed PhAIL as both a benchmark and proof point for its broader infrastructure layer covering data pipelines, unified inference, and evaluation across model vendors and hardware.
Source - 2025-09-25 - Company Formation - Positronic Robotics publicly launches to build infrastructure for ML robotics
Positronic Robotics publicly emerged from stealth and announced its focus on making professional-grade ML robotics more approachable. The company introduced its early runtime, dataset tooling, and collect-to-validate workflow for real-world robot learning. The launch positioned Positronic around infrastructure for end-to-end ML robotics rather than robot hardware.
Source