The Desktop IDE for Edge AI
Fusion Studio brings model training, optimization, benchmarking, and deployment into one embedded AI workflow eliminating fragmented toolchains and reducing the time between proof of concept and production deployment.
Edge AI development is fragmented.
Fusion Studio fixes that
The fragmented toolchain
Shipping a model to the edge today means stitching together five separate tools: a training framework, a quantization script, a compiler, a benchmarking harness, and a deployment pipeline. Each one has its own file format, version constraints, and failure modes. The work isn't the model, the work is the integration.
The wrong fix
The industry's response has been to hide the complexity behind AI wizards and one-click "build my model for me" buttons. That's the wrong fix. Embedded engineers don't want a magic box that guesses at their architecture, their quantization strategy, and their memory budget. They want a real IDE - one that exposes every step of the pipeline, gives them control over every decision, and ships AutoML as a tool, not as the product.
Fusion Studio is that IDE
You write the model. You set your training parameters. We handle the toolchain. Fusion Studio exposes every step of the pipeline with full transparency - no magic boxes, no guesswork. Real control for the engineers building game-changing products
Edge AI development is fragmented. Fusion Studio fixes that
Shipping a model to the edge today means stitching together five separate tools: a training framework, a quantization script, a compiler, a benchmarking harness, and a deployment pipeline. Each one has its own file format, version constraints, and failure modes. The work isn’t the model – the work is the integration.
The industry’s response has been to hide the complexity behind AI wizards and one-click “build my model for me” buttons. That’s the wrong fix. Embedded engineers don’t want a magic box that guesses at their architecture, their quantization strategy, and their memory budget. They want a real IDE – one that exposes every step of the pipeline, gives them control over every decision, and ships AutoML as a tool, not as the product.
Fusion Studio is that IDE. You write the model. You own the weights. We handle the toolchain.
- Trusted by teams at
Run on the silicon you already ship
built around the toolchains and frameworks that define the embedded silicon ecosystem – with the deepest integration on Arm.
“[Fusion Studio] is reducing complexity, accelerating innovation, and supporting teams to move to production more efficiently.”
John Thompson
Senior Director of Edge AI Software, Arm
Run on the Silicon you already have
Fusion Studio targets the silicon embedded teams ship into production today, with toolchain integrations validated on real hardware – not just simulators
Arm Ethos-U85
- Core: Ethos-U85 NPU
- Toolchain: ExecuTorch, Vela and TFLite
Arm Ethos-U55
- Core: Ethos-U55 NPU
- Toolchain: ExecuTorch, Vela and TFLite
Alif Ensemble Series
- Core: Ethos-U55/U85 + Cortex-M55
- Toolchain: ExecuTorch, Vela and TFLite
STMicroelectronics STM32N6
- Core: Neural-ART Accelerator
- Toolchain: TFLite
- coming soon
Infineon PSoC Edge E84
- Core: Infineon PSoC Edge E84
- Toolchain: ExecuTorch & TFLite
- coming soon
Run on the Silicon you already have
Fusion Studio targets the silicon embedded teams ship into production today, with toolchain integrations validated on real hardware – not just simulators
| Silicon Platform | Core / Accelerator | Toolchain | Status |
|---|---|---|---|
| Arm Ethos-U85 | Ethos-U85 NPU | ExecuTorch · Vela · TFLite | Production |
| Arm Ethos-U55 | Ethos-U55 NPU | ExecuTorch · Vela · TFLite | Production |
| Alif Ensemble E3/E7/E8 | Ethos-U55/U85 + Cortex-M55 | ExecuTorch · Vela · TFLite | Production |
| STMicroelectronics STM32N6 | Neural-ART Accelerator | TFLite | Beta |
| Infineon PSoC Edge E84 | Ethos-U55 + Cortex-M55 | TFLite · ExecuTorch | Beta |
Working with silicon outside this list?
Contact our engineering team – let’s discuss platform integration scope and timeline.
Your data stays on your machine
Fusion Studio is a desktop application. Training datasets, model weights, compiled binaries, and benchmark results live on your hardware.
For security, defense, government, and industrial teams working with sensitive datasets, this isn’t a preference – it’s an architectural requirement, and one Fusion Studio is built to meet. Once installed, the application has no runtime dependency on cloud services for training, optimization, compilation, or benchmarking. Fusion Studio connects to the internet only to download Zoo models you explicitly select, and to check for application updates you can disable.
When Standard Won't fit:
Some deployment environments need more than a standard install – fully air-gapped networks, classified compute boundaries, custom toolchain integrations, hardened security configurations, on-prem build pipelines.
Fusion Studio is backed by a ~400-engineer embedded software team with 20+ years of shipping production systems. If your environment required your product, configured for our requirements, we have the engineering capacity to build it.
Pure-product vendors point you at their roadmap. We point you at our engineering team.
Check out Fusion Studio today
Fusion Studio Explorer is free to download – full IDE, full Zoo access, no credit card, no time limit. Upgrade to Builder or Fleet when your team needs advanced AutoML, on-prem hosting, or dedicated deployment support.