ML Model Orchestration Engine for Data Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The lack of standardization in machine-learning (ML) models from disparate sources complicates data transfer between models, requiring time-consuming normalization processes, and users often need to access multiple service providers for specific functionalities.
Innovation Solution
A system and method for managing and deploying pre-trained ML models through an orchestration system that selects and integrates models based on client demands, using a model orchestration engine and database to build execution pipelines, normalizing data inputs and outputs across models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If ML models from disparate sources are used to provide specific functionalities, then the capability and versatility of the system is improved, but the complexity of data integration and normalization increases
Solution Approach 1:
The patent introduces an orchestration system as an intermediary layer between disparate ML models and users. This orchestration system manages model selection, data routing, and normalization processes, thereby enabling access to diverse ML functionalities while abstracting away the integration complexity from end users.
Solution Approach 2:
The orchestration system provides universal data specifications and standardized interfaces that work across multiple different ML models from various sources. This allows a single standardized interface to interact with diverse models, reducing the need for model-specific integration code and simplifying the overall system architecture.
2Ease of operation
If data normalization processes are implemented to enable data transfer between ML models, then the compatibility and ease of operation is improved, but the processing time and loss of time increases
Solution Approach 1:
The system performs data normalization and formatting operations in advance, before data is passed between ML models. By pre-processing data to match expected specifications upfront, the system eliminates the need for repeated normalization operations during model execution, thereby reducing overall processing time while maintaining compatibility.
Solution Approach 2:
The orchestration system dynamically adjusts data parameters and formats based on the specific requirements of each ML model in the execution pipeline. By automatically transforming data parameters to match model expectations, the system achieves compatibility without requiring manual intervention or excessive processing time.
3Adaptability or versatility
If multiple service providers are accessed to obtain ML models with specific capabilities, then the functional completeness is improved, but the device complexity and difficulty of operation increases
Solution Approach 1:
The orchestration system merges access to multiple ML models from different service providers into a single unified interface. Users interact with one standardized system rather than multiple separate provider interfaces, while the orchestration system internally manages connections to diverse model sources, thereby maintaining functional completeness while simplifying operations.
Data Source
AI summary
Disclosed herein are systems and methods for managing and deploying pre-trained machine -learning (ML) models from disparate third-party systems and a host system. An orchestration system selects effective pre-trained ML models to suit a client user’s business operation demands and data analysis requirements. A model orchestration engine accesses and maintains a model orchestration database containing a catalogue of ML models. The catalog indicates, for example, the capabilities, data inputs, and data outputs of each ML model. The orchestration engine builds an execution pipeline of certain ML models according to a client request for an operation or function and the client data submitted from a client device. The orchestration server can format or normalize the client data or output data received from the client device or an earlier model according to a data specification, thereby generating input data for subsequent models in the pipeline.