Funding
Funding not disclosed
Founders
Product
Problem
Deploying and managing machine learning models on edge devices, such as smartphones, is complex due to the diversity of hardware, operating systems, and runtime environments. Converting and optimizing models for each specific target environment is time-consuming and requires specialized expertise. Furthermore, distributing and updating models across a fleet of devices can be challenging, leading to delays and inefficiencies.
Solution
Makinas provides an MLOps platform designed to simplify the deployment, optimization, and orchestration of machine learning models on edge devices. The platform allows users to convert models from various frameworks, including TensorFlow, PyTorch, and CoreML, into multiple runtime formats like TF Lite, TensorRT, and TVM in a single step. It functions as a content delivery network (CDN) for models, enabling them to be easily served across a diverse device ecosystem and updated without requiring full software updates. This allows developers to map models to specific devices, platforms, applications, and users, ensuring efficient and targeted deployment.
Target Audience
The primary target audience includes machine learning engineers, mobile app developers, and enterprises seeking to deploy and manage AI models on edge devices, such as smartphones and embedded systems.
Features
- Model conversion from TensorFlow, CoreML, and PyTorch to TF Lite, TensorRT, CoreML, and TVM.
- Model optimization for target runtime environments.
- Content delivery network (CDN) functionality for model distribution.
- Device and application mapping for targeted model deployment.
- Secure model storage with options for public or private access.
- Support for community model sharing and prototyping.