embedUR systems and Alif Semiconductor have formed a partnership to add native support for Alif’s Ensemble Series MCUs to embedUR’s ModelNova Fusion Studio, enabling a complete train-to-deploy workflow on microcontrollers with hardware-accelerated transformer support.

The collaboration launches an end-to-end Edge AI platform for microcontrollers, integrating Ensemble Series with ModelNova Fusion Studio to enable on-device train-to-deploy workflows without relying on cloud tools.

The two companies announced a joint effort to bring Ensemble Series microcontrollers into Fusion Studio, delivering a seamless train-to-deploy workflow on desktop for MCUs with on-device transformer acceleration.

A joint technical webinar is set for March 4, 2026 to demonstrate the end-to-end workflow from dataset preparation to on-device deployment, with registration details provided by the companies.

The webinar will showcase the full pipeline, from dataset preparation to deployment on hardware, and registration can be completed online.

At embedded world 2026 (March 10–12 in Nuremberg), live demonstrations will feature Fusion Studio deploying Edge AI models on Ensemble hardware in real time.

Demonstrations at embedded world will highlight Fusion Studio’s real-time deployment of Edge AI models on Ensemble hardware at the event venue.

Executives emphasize that Fusion Studio eliminates cloud costs and toolchain fragmentation by offering a complete workflow from unlabeled data to production hardware, accelerating time-to-market.

Industry leaders frame the collaboration as providing a cloud-free, end-to-end path from data to on-device inference, aligning with Ensemble’s edge capabilities.

Fusion Studio’s Ensemble support targets computer vision tasks such as object detection, pose estimation, and image segmentation, with additional AI modalities planned for 2026–2027.

Alif’s Ensemble Series features second-generation E4, E6, and E8 MCUs with over 450 GOPS and substantial on-chip memory, enabling on-device execution of small language models and transformer-based vision models.