Mellerikat MLOps

mellerikat MLOps is a tool that enables continuous operation and optimization of AI/ML models. It supports the entire process of AI/ML model development, deployment, and operation, allowing users to easily manage AI models without requiring professional knowledge of AI/ML.

Why MLOps?

Why is it difficult to operate AI/ML technology as a service?

It is easy to solve problems as one-off projects with AI/ML, but applying them to actual service operation environments is not easy. In service operation, you need to prepare things like connecting data pipelines for inference and a structure that allows ML models to be deployed in production. In addition, continuous operation of models requires features such as monitoring model performance, detecting and diagnosing performance degradation, and retraining and redeploying models. mellerikat MLOps supports the service operation of AI/ML models.

Service Architecture

Easy and Sustainable MLOps

mellerikat MLOps is divided into AI model development and operation domains. In the AI model development domain, data scientists develop AI Solutions based on ALO and register them in AI Conductor. In the AI model operation domain, field personnel download AI Solutions from AI Conductor via Edge Conductor, train AI models, and deploy them to the Edge. Field personnel can monitor model performance and update models through retraining and redeployment, even without professional knowledge of AI or ML.

AI Model Development

Industrial-proven AI Contents

mellerikat provides AI Contents whose technology and effectiveness have already been proven in various industries. Users can download these verified AI Contents and easily customize them for their own problem situations with minimal modification. This allows for quick and easy development and registration of AI Solutions.

Framework for AI/ML Model Development, ALO

ALO is a framework that converts AI/ML algorithm code developed by developers into deployable AI Solutions. AI Solutions registered through the ALO framework can utilize mellerikat's MLOps solution to easily perform the entire process from model training to deployment in the service environment via a web UI.
This allows users to efficiently develop, train, optimize, and deploy AI models without complex processes.

Systematic AI Solution/Model Management, AI Conductor

AI Solutions developed through ALO are registered in AI Conductor. Users can manage AI Solutions in operation by version using AI Conductor. In addition, AI Conductor allows you to manage model training resources and training history required for MLOps. AI Conductor serves as a hub for AI Solutions and AI Models, supporting users' continuous MLOps operations.

Innovative AI Model Update Process

To continuously operate AI models, their performance must be consistently maintained or improved. mellerikat MLOps continuously monitors AI model performance, detects performance degradation, and provides features to update AI models with the latest technology. By simply updating the AI Solution in operation with the latest AI/ML technology, you can quickly and easily replace the running AI model with the latest one. Experience continuous optimization of AI models using mellerikat's innovative AI Solution and AI Model deployment process.

Competitive Model Advancement through Competitions

To support continuous model performance improvement, mellerikat provides a Competition platform feature. Through the Competition platform, AI/ML experts can propose even better AI Solutions or AI Models using their own ideas and technologies. Host AI Competitions to experiment with AI models reflecting the latest AI/ML trends and experience continuous model advancement.

AI Model Operation

Field Expert-Centric MLOps Support System, Edge Conductor

With an easy and intuitive user interface, even field experts without professional knowledge of AI/ML can easily train, deploy, and operate models.

Innovative AI Model Management/Deployment Structure

Deploy AI models to numerous endpoints and utilize them at the edge. As long as you install Edge App on each device, you can freely deploy and use AI models, and easily replace them. Deploy AI models at scale and experience the expansion of AI with Edge Conductor.

Scalable Edge AI Operation/Monitoring

With Edge Conductor, you can manage numerous endpoints, monitor inference results of deployed AI models in real time, and collect inference results.

Model Optimization through Data Re-labeling & Re-training

mellerikat MLOps provides features to monitor AI model inference results in real time and continuously improve model performance. When operating AI/ML models for a long time, model performance on the latest data may degrade, requiring retraining and redeployment. You can create datasets from inference results collected by Edge Conductor and build high-quality training datasets by performing data re-labeling using domain knowledge. All these processes can be easily performed through the intuitive UI of Edge Conductor, even without professional knowledge of AI/ML.

Build Edge AI with Just Edge App Installation

Edge App is software used together with Edge Conductor to easily deploy and manage AI models. Edge App can run on various operating systems and automatically configures the environment for running AI models. A device with Edge App installed transforms into an Edge AI Device capable of running AI models. You can deploy AI models via Edge Conductor and run them on devices with Edge App installed. Edge App provides features to monitor inference results in real time and continuously improve model performance.

Components

Complete MLOps Solution

mellerikat consists of five technical components (Edge App, Edge Conductor, AI Conductor, AI Learning Organizer, AI Contents).
To solve the high technical barriers and continuous operation difficulties of MLOps, mellerikat separates the traditional MLOps domain and Model as a Service domain, providing optimized features for each area.