morfics
Products
Ready AI
Vision, LLM, embeddings, speech and published models.
AI Cloud
Run accelerated workloads on managed or private infrastructure.
Web AI Apps
Frontend + backend + accelerated AI in one product workflow.
Embedded
Deploy intelligence into devices, robotics and physical systems.
Advanced engineering
Design exploration, custom architectures, toolchains, runtimes and low-level control.
HubHardwareDevelopersEnterprise
Start
FPGAI compiler

Connect AI models with reusable hardware implementations.

FPGAI is the open compiler layer underneath Morfics. It maps inference and training workloads to FPGA systems and will import implementation packages published by the community across AMD/Xilinx, Intel, Lattice, open-source, and custom flows.

Open ONNX StudioBrowse implementations

Import

Start from ONNX, an FPGAI project, or a community implementation package.

Analyze

Inspect operators, tensor movement, precision, memory, interfaces, and target compatibility.

Select implementations

Use built-in kernels or reusable HLS, VHDL, Verilog, and SystemVerilog packages.

Build and validate

Generate target-specific projects, reports, testbenches, runtime artifacts, and deployment bundles.

Open extension contract

The community can add more than models.

FPGAI registries are planned for operators, complete models, backends, boards, memory policies, transport, optimizers, losses, datasets, reports, and runtime adapters.

HLS C++
VHDL
Verilog / SystemVerilog
FPGAI-native packages
AMD/Xilinx
Intel FPGA
Lattice
Custom boards and toolchains

Scientific IR

Explicit tensor, memory, transport, precision, inference, and training semantics.

Plugin architecture

Contributors add packages through stable manifests and registries instead of editing central compiler files.

Hardware path

Target user-owned boards, managed accelerator infrastructure, and private or embedded deployments.

Compatibility path

See how community assets can become compiler-ready implementations.

The flow explains native packages, metadata-compatible assets, adapter requirements, and future deployment targets.

FPGAI Compiler

One path from source to hardware

Pick an input format to trace its current compatibility path.

  1. Community package

    Your source, published as a package.

  2. Morfics metadata

    Normalized description of the source.

  3. FPGAI registry

    Matched against known implementations.

  4. Metadata import

    Metadata compatible

  5. Deployment target

    PYNQ-Z2 · KV260 · Alveo U50 / U55C

ONNX · Metadata compatible

Operators are matched against the registry; supported ops map directly, the rest are flagged.

Explore the compiler