Dynamic AI Model and Hardware Routing for Efficient Task Execution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional systems are inflexible and computationally inefficient when utilizing artificial intelligence models and hardware environments, leading to excessive bandwidth usage, computational inefficiencies, and compromised output quality due to limitations in model and environment selection.

Innovation Solution

The intelligent selection and execution platform dynamically selects optimal machine-learning models and hardware environments based on workload features and task routing metrics, allowing for fallback options and intelligent scheduling of tasks based on bandwidth availability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If conventional systems utilize a single artificial intelligence model or a small handful of models, then the system complexity is reduced and ease of operation is improved, but computational efficiency deteriorates and bandwidth usage increases

Engineering Contradiction:

Improveease of operationVSAvoidcomputational efficiency

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The system implements a universal task routing layer that can dynamically direct tasks to multiple different AI models based on task characteristics. This routing mechanism allows a single system architecture to support multiple models without requiring separate systems for each model, achieving multi-functionality while maintaining operational simplicity.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system dynamically selects which AI model to use for each task based on real-time analysis of task complexity, model availability, and performance metrics. This dynamic selection process optimizes computational efficiency by matching the right model to the right task, preventing both underutilization and overutilization of computational resources.

Inventive Principle:

Principle #15Dynamics

2Manufacturing precision

If conventional systems execute tasks on large artificial intelligence models, then output quality is improved, but computational power consumption increases and processing speed decreases

Engineering Contradiction:

Improveoutput qualityVSAvoidcomputational power

Core Design Contradiction:

Manufacturing precisionVSPower

Solution Approach 1:

The system applies local quality by matching specific task requirements with corresponding model capabilities. Instead of uniformly using large models for all tasks, the routing mechanism analyzes task characteristics and directs simple tasks to smaller, more efficient models while reserving large models for complex tasks that truly require their computational power, thus optimizing the quality-power tradeoff locally for each task.

Inventive Principle:

Principle #3Local quality

3Speed

If conventional systems utilize local hardware environment for task execution, then response time is improved, but system reliability deteriorates when local resources are exhausted or offline

Engineering Contradiction:

Improveresponse timeVSAvoidsystem reliability

Core Design Contradiction:

SpeedVSReliability

Solution Approach 1:

The system implements beforehand cushioning by pre-configuring fallback hardware environments and establishing routing rules that automatically activate when local resources are exhausted or offline. This preparatory measure ensures continuity of service by having alternative execution paths ready in advance, preventing system failures when local resources become unavailable.

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

4Speed

If conventional systems initiate training tasks on local hardware environment, then training speed is improved, but bandwidth availability for other tasks is reduced and resource allocation efficiency deteriorates

Engineering Contradiction:

Improvetraining speedVSAvoidresource allocation efficiency

Core Design Contradiction:

SpeedVSProductivity

Solution Approach 1:

The system implements periodic action by scheduling training tasks during periods of low bandwidth utilization and normal task execution during peak periods. The routing mechanism monitors bandwidth availability and dynamically adjusts when to initiate training tasks versus when to prioritize inference tasks, creating a periodic pattern that optimizes both training speed and overall resource allocation efficiency.

Inventive Principle:

Principle #19Periodic action

Data Source

PatentUS12386667B2Dynamically selecting artificial intelligence models and hardware environments to execute tasks

Publication Date: 2025.08.12 DROPBOX INC

AI summary

The present disclosure relates to systems, non-transitory computer-readable media, and methods for selecting machine-learning models and hardware environments for executing a task. In particular, in one or more embodiments, the disclosed systems select a designated machine-learning model for executing a task based on workload features of the task and task routing metrics for a plurality of machine-learning models. In addition, in one or more embodiments, the disclosed systems select a designated hardware environment for executing the task based on workload features for the task and task routing metrics for a plurality of hardware environments. In some embodiments, the disclosed systems select a fallback machine-learning model and a fallback hardware environment for executing the task if the designated machine-learning model or designated hardware environment are unavailable. Moreover, in one or more embodiments, the disclosed systems can pause and initiate tasks based on bandwidth availability.