Thesis Markets Episodes Community Blog
Phasor · NeuroAI for autonomous machines

AI uses too much energy. Nature has already solved it.

We build Neuro-based models for autonomy in difficult places. Bio-inspired perception and navigation for spacecraft, defense aircraft, and self-driving cars. Built to run on watts where conventional AI burns kilowatts.

See the demo →
Backed by
NVIDIA
Inception
Open Neuromorphic
community
Founder’s Inc
Canopy 2026
Anthropic
for startups
01 — What we build

Building autonomy for machines,
the way nature built it.

A dragonfly catches prey mid-flight on a brain smaller than a grain of rice. A human reasons about the world on twenty watts. Today’s AI systems burn 2,500 but can’t see through a midnight blizzard, a brownout, or a lunar shadow.

The problem isn’t more compute. It’s the wrong primitive. Today’s AI samples the world and brute-forces pattern recognition with GPUs. Biology reacts to change, spikes when it matters, and ignores the rest.

We build Neuro-based models for autonomy in difficult places; spiking neural networks trained on event-based data. Currently, we are focused on Spatial Reasoning for perception and navigation in the next generation of embodied machines.

02 — SPACES

Same brain.
Three frontiers.

SPACE
Lunar lighting conditions.

NASA explicitly lists neuromorphic cameras as a need. Shortfall ID 1601.

DEFENSE
Degraded Visual Environment.

The Pentagon’s term. 75% of arid-theater helicopter mishaps.

AUTOMOTIVE
Low-visibility events.

NHTSA EA26002, October 2024. Tesla FSD collisions in sun glare, fog, and dust.

Same model across all three. Bolts on as a co-processor to whatever certified compute the platform already flies: RAD750, INTEGRITY-178, NVIDIA Drive, Mobileye EyeQ.

03 — Episodes

The world’s largest event-based dataset.
and getting larger

Episodes is the data layer underneath Phasor. We gave 652 event-based datasets (and growing) collected and structured for models that learn from change instead of frames. It’s how we train NeuroAI.

It’s also a live research surface. Used by one frontier research lab and four academic groups, all training their own models on it. If you’re researching neuromorphic perception, navigation, or sensorimotor control, this is the data you’ve been looking for.

Explore the dataset →
A note from us

If you’re researching neuromorphic perception or navigation, tell us what data you need. We’ll get it to you.

Email v eric at project phasor dot com, or drop a request.

04 — Community

Building in the open.
Come along.

Watch what we’re building. Join the conversation. Whether you’re a researcher, an engineer, or just curious about what’s next for AI, we’d love to have you.

Join us on Discord → Follow on LinkedIn →

Phasor

NeuroAI for autonomous machines. Bio-inspired perception and navigation for spacecraft, defense aircraft, and self-driving cars. Built to run on watts where conventional AI burns kilowatts.

Site

Backed by

Thesis

A dragonfly catches prey mid-flight on a brain smaller than a grain of rice. A human reasons about the world on twenty watts. Today’s AI systems burn 2,500 — and still can’t see through a midnight blizzard, a brownout, or a lunar shadow.

The problem isn’t more compute. It’s the wrong primitive. Today’s AI samples the world on a clock and brute-forces pattern recognition with GPUs. Biology reacts to change, spikes when it matters, and ignores the rest.

We build NeuroAI — spiking neural networks trained on event-based data. Watts instead of kilowatts. Microseconds instead of milliseconds. The brain version of perception and navigation, for the next generation of embodied machines.

What we build

Phasor is the model layer for embodied intelligence — perception and navigation, trained on event-based data, running on whichever neuromorphic chip wins. Spiking neural networks doing the work GPU-trained CNNs cannot: real-time perception in degraded conditions, on a watts-scale power budget, with reflex-class latency.

Markets

Space — Lunar lighting conditions

NASA explicitly lists neuromorphic cameras as a need. Shortfall ID 1601.

Defense — Degraded Visual Environment

The Pentagon’s term. 75% of arid-theater helicopter mishaps.

Automotive — Low-visibility events

NHTSA EA26002, October 2024. Tesla FSD collisions in sun glare, fog, and dust.

Same model across all three. Bolts on as a co-processor to whatever certified compute the platform already flies — RAD750, INTEGRITY-178, NVIDIA Drive, Mobileye EyeQ. Months to integrate, not years to re-certify.

Episodes

Episodes is the data layer underneath Phasor — 652 event-based datasets, collected and structured for models that learn from change instead of frames. It’s how we train NeuroAI.

It’s also a live research surface. Used by one frontier research lab and four academic groups, all training their own models on it. If you’re researching neuromorphic perception, navigation, or sensorimotor control, this is the data you’ve been looking for.

Community

Contact

Vision

Every machine — the car driving you home through a midnight blizzard, the satellite making first contact with a new planet — should think with the efficiency of a brain.

Meta

Copyright © 2026 Phasor. Press / to toggle MACHINE mode. Press Esc to return to HUMAN.