One architecture, any signal.
A single analog compute core, in two products: AML100 shipping today across sensor modalities, and AML200 extending analog AI to radio frequency.
AML200.
The first analog AI chip to classify gigahertz RF signals before the ADC. AML200 sits between the antenna array and the RF SoC, cutting compute and power across the signal chain.
The same capability serves defense and commercial RF sensing where digital is simply too inefficient and slow.
AML100.
The world's only commercially available fully analog AI compute core — a production IC with a Python SDK, deployable today across vibration, acoustic, current, bio, and other <300 kHz sensor modalities.
Always-on inference at under 20µA, entirely on-device, programmed in Python and PyTorch with no analog or firmware expertise required.
MARC100.
The networked sensor node for distributed application deployments. Built around the AML100.
MARC100 Sensor Platform
A compact wireless sensor device built around the AML100 — the multi-node front end for distributed monitoring and application deployments.
A Python SDK that fits your ML workflow.
Build your model in TensorFlow or PyTorch, construct and measure AnalogML signal chains in AnalogML™ Connect, then compile to a bytestream for the chip.
Aspinity's software development kit integrates with standard machine-learning tools, so your team can develop, test, and verify AnalogML models for the AML100 without learning analog design.
Build your own detection algorithms for acoustic, piezo, vibration, and other sensor events — or start from Aspinity's purpose-built event-detection models. The core is field-programmable and interfaces with up to four analog sensors for sensor-fusion accuracy.
What's in the SDK
- DevelopDefine and train AnalogML configurations using familiar ML tooling.
- VerifyTest and validate model behavior against your captured signals before deployment.
- DeployProgram the AML100 as an always-on analog front end that wakes the MCU only on an event.
Ready to evaluate the silicon?
Tell us your signal, your power budget, and your timeline — we'll help you figure out the fit.