Skip to content

matlab/matlab-agentic-toolkit / skills-catalog/ai-and-statistics/matlab-deploy-embedded-ai

matlab-deploy-embedded-ai

Verified

Deploy AI models to embedded hardware using MathWorks tools (MATLAB, Simulink, Embedded Coder). Covers two workflow patterns: (1) MathWorks-native or imported models rebuilt as dlnetwork for lean hardware, (2) direct C/C++ code generation from PyTorch and LiteRT models. Both patterns support all targets (Cortex-M/A/R, x86, GPU). Trigger when: user wants to deploy AI to embedded targets; generate C/CUDA from neural networks; compress AI models for MCU; integrate AI in Simulink for system-level simulation; import PyTorch/ONNX/TensorFlow models for embedded deployment; optimize AI for resource-constrained hardware; or use loadPyTorchExportedProgram, loadLiteRTModel, importNetworkFromPyTorch, importNetworkFromONNX, importNetworkFromTensorFlow, importNetworkFromKeras, dlquantizer, exportNetworkToSimulink, or Embedded Coder with AI models.

references/shared/environment-setup.md

View on GitHub
# Environment Discovery (Silent)

Run this phase silently at the start of every session. Do NOT prompt the user.

## Step 1: Detect Toolboxes

Use `detect_matlab_toolboxes` MCP tool to enumerate installed toolboxes and MATLAB version.

Record availability of:
- Deep Learning Toolbox (DLT)
- Statistics and Machine Learning Toolbox (Stats/ML)
- MATLAB Coder
- Simulink Coder
- Embedded Coder
- GPU Coder
- HDL Coder
- Deep Learning HDL Toolbox
- Fixed-Point Designer
- Simulink
- Signal Processing Toolbox
- Image Processing Toolbox
- Reinforcement Learning Toolbox

**Gate:** STOP if **both** Deep Learning Toolbox and Statistics/ML Toolbox are missing.

## Step 2: Detect Support Packages

Support packages are **NOT** returned by `detect_matlab_toolboxes`. Query separately:

```matlab
pkgs = matlabshared.supportpkg.getInstalled;
if ~isempty(pkgs)
    disp({pkgs.Name}');
end
```

Check for these support packages:

| Support Package | Needed For |
|----------------|-----------|
| Deep Learning Toolbox Converter for PyTorch Models | Pattern 1 import from PyTorch |
| Deep Learning Toolbox Converter for ONNX Model Format | Pattern 1 import from ONNX |
| Deep Learning Toolbox Converter for TensorFlow Models | Pattern 1 import from TensorFlow |
| Deep Learning Toolbox Model Compression Library | Pattern 1 compression (pruning, projection) |
| AI Verification Library for Deep Learning Toolbox | AI verification (OOD detection) |
| Deep Learning Support from MATLAB Coder | Pattern 1 code generation (`coder.loadDeepLearningNetwork`) |
| MATLAB Coder Support Package for PyTorch and LiteRT Models | Pattern 2 (direct PyTorch/LiteRT to C) |
| Embedded Coder Support Package for ARM Cortex-M Processors | CMSIS/CMSIS-NN optimized code for Cortex-M targets |

**Detection pattern:**
```matlab
pkgs = matlabshared.supportpkg.getInstalled;
hasPyTorchConverter = ~isempty(pkgs) && any(contains({pkgs.Name}, "Converter for PyTorch"));
hasONNXConverter    = ~isempty(pkgs) && any(contains({pkgs.Name}, "ONNX"));
hasPyTorchSPKG     = ~isempty(pkgs) && any(contains({pkgs.Name}, "PyTorch and LiteRT"));
hasCompLib         = ~isempty(pkgs) && any(contains({pkgs.Name}, "Model Compression"));
hasAIVerif         = ~isempty(pkgs) && any(contains({pkgs.Name}, "AI Verification"));
hasArmCortexM      = ~isempty(pkgs) && any(contains({pkgs.Name}, "ARM Cortex-M"));
```

## Step 3: Check Python Environment (if PyTorch workflows possible)

```matlab
pe = pyenv;
disp(pe.Version);    % Should be 3.9+
disp(pe.Executable); % Path to Python
```

If Python is not configured, set it with the path to your Python executable:
- Windows: `pyenv('Version', 'C:\Users\<user>\AppData\Local\Programs\Python\Python311\python.exe')`
- macOS/Linux: `pyenv('Version', '/usr/local/bin/python3')`

For Pattern 2: Python environment must have `torch >= 2.0`.

If local agent instructions or an `AGENTS.md` file specify a Python environment,
follow those instructions. On MATLAB installations with the PyTorch converter
support package, a read-only Python 3.11 environment with PyTorch may be available
under the support package root:

```matlab
spRoot = matlabshared.supportpkg.getSupportPackageRoot;
pyExe = fullfile(spRoot, '3P.instrset', ...
    'pythonstandalone_3.11.instrset', 'win64', 'python', 'python.exe');
torchRoot = fullfile(spRoot, '3P.instrset', ...
    'pytorch_mlir_python_3.11.instrset', 'win64');
```

Use this bundled Python for PyTorch reference-vector generation or `.pt2` export
when the project instructions prefer it. Do not install packages into this
support-package Python; it is managed by MATLAB.

### Python/MATLAB Interop Pitfall: Numpy Array Dimensions

When loading numpy `.npy` files into MATLAB via `py.numpy.load` or `single(npArray)`:
- A 1-D numpy array of shape `(N,)` becomes a MATLAB **row vector** `[1 × N]`
- This causes implicit expansion bugs when subtracted from column vectors `[N × 1]`

**Always force column vectors after loading numpy data:**
```matlab
refOutputs = single(py.numpy.load("outputs.npy"));
refOutputs = refOutputs(:);  % Force column vector
```

## Step 4: Present Environment Summary

Present a concise summary to the user:

```
Environment Summary:
- MATLAB R20XXx
- Toolboxes: [list available]
- Support packages: [list available]
- Missing for your workflow: [list if any]
- Python: [version] at [path] (or "not configured")
```

If a required support package is missing, inform the user:
"[Package name] is not installed. Please download it from the Add-On Explorer
(Home > Add-Ons > Get Add-Ons) and let me know when it's ready."

**Do NOT install on the user's behalf. Wait for confirmation.**


Copyright 2026 The MathWorks, Inc.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 321kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 197kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall