Home Compatibility

Compatibility

 

Compatibility
 

Built to Fit. Ready to Deploy.

 

 OptoML AI processors are designed for seamless integration — supporting standard frameworks, interfaces, and tools used by ML and embedded developers today.
 

Model & Framework Compatibility

 

 Our architecture supports quantized AI models optimized for ultra-low-power inference. We provide tools and reference flows for adapting standard AI models to run on OptoML hardware.
 

Model Types Supported

 

  • Convolutional Neural Networks (CNNs)

  • Fully Convolutional Networks (e.g., U-Net)

  • Lightweight Transformer variants (planned)

  • Quantized models (INT4, INT8)

  • Temporal and frame-based inference
 

Framework Interoperability

 

  • PyTorch (quantized export)

  • TensorFlow Lite (quantized models)

  • ONNX format (preferred import route)

  • OptoML SDK (custom quantizer & converter toolchain)
 

Hardware & Interface Support

 

 OptoML chips are built with practical deployment in mind — compatible with widely used interfaces and deployment stacks.
 

I/O & Peripheral Support

 

  • SPI / I2C / UART

  • MIPI CSI (for camera inputs)

  • GPIO-based trigger and interrupts
 

Host Interface Options

 

  • MCU (ARM Cortex-M / RISC-V)

  • Linux-based hosts (via SPI bridge)

  • Optional direct mode (standalone)
 

Form Factors

 

  • SoC embedded

  • Optional AI co-processor board (QFN/BGA-based)
 

Developer Tools & SDK

 

 We offer a lightweight, developer-friendly toolchain to convert and deploy AI models onto OptoML silicon.
 

OptoML SDK includes

 

  • Model converter (ONNX → OptoML format)

  • Quantizer for INT4/INT8 compression

  • Debugger and power profiler tools

  • Simulation tools for FPGA/test-chip mode

  • API for host communication and event triggering
 

Supported Environments

 

  • Linux, Windows (toolchain)

  • Embedded RTOS (host control)

  • Python CLI interface (conversion, profiling)
 

Security Compatibility

 

Hardware hooks for secure integration:

  • Encrypted model loading

  • Hardware-based secure boot

  • Configurable access control