Agentic workflows are AI-powered software systems that chain multiple models and external tools to tackle complex tasks, such as analyzing a video and answering questions about it. However, the design and deployment of these fragmented systems often lead to inefficiencies, resulting in wasted computation, energy, and costs. To enhance efficiency, researchers from MIT and Microsoft developed an intelligent system that streamlines the design process of agentic workflows and automatically optimizes their implementation.
With this new method, developers can describe what they want the agentic workflow to accomplish in plain language, without needing to specify all application details in advance. The system automatically determines the best models and tools to use, as well as the ideal hardware configuration and computational resource allocation when the workflow is executed by a cloud provider. It adjusts those configurations on the fly based on user priorities, such as minimizing costs or maximizing speed. When tested on various agentic workloads, this new system reduced the number of computational units needed for deployment, significantly cutting energy requirements and costs compared to traditional approaches without compromising performance.
"Agentic workflows are becoming very complicated and quickly becoming the backbone of what cloud providers are doing. Energy usage is a huge concern, so we need to be very careful about how efficient these workflows are. It is very easy to over-allocate resources, wasting energy and money. Enabling a cloud provider to intelligently make these workflows more resource-optimal is a win for everyone involved," says Gohar Chaudhry, an EECS graduate student and lead author of a paper on this system.
The system, named Murakkab (an Urdu word meaning a composition of things), aims to optimize the entire agentic workflow process. Murakkab allows developers to create an agentic workflow by describing their intent for the application in high-level terms rather than detailing how the many components should be combined. For instance, a developer might describe a video Q&A application that extracts key frames, generates a transcript, and answers user queries about the video. Murakkab automatically identifies the best existing models and tools to assemble into the workflow and determines which components need to run sequentially and which can run in parallel to enhance performance.
When the cloud provider deploys the application for a customer, Murakkab optimizes the workflow by configuring its components to meet user constraints, such as prioritizing accuracy while meeting latency requirements. It adaptively identifies ideal hardware allocations and deployment schedules to maximize efficiency in real time, then generates a workflow ready for the cloud provider to execute. In tests with diverse agentic workflows for video Q&A and code generation, Murakkab met user requirements while using only about 35 percent of the computation required by other methods. It consumed only about 27 percent as much energy for less than 25 percent of the cost. The dynamic nature of Murakkab also allows users to balance tradeoffs. In one case, the system lowered energy consumption of an agentic workflow by more than an order of magnitude with only about a 2 percent drop in accuracy for the customer.
Next, the researchers plan to expand their system to more complex workflows and larger computing clusters while exploring opportunities to optimize new agentic applications. "There is a lot of potential to make these workflows more resource-optimal so they consume far less energy, but we need to be thinking about this at the scale of major cloud platforms," says Chaudhry.
This research was supported, in part, by the Semiconductor Research Corporation and the U.S. Defense Advanced Research Projects Agency.
Blogger's Review: The emergence of the Murakkab system marks a significant advancement in AI agent workflows, achieving dynamic optimization of configurations that not only enhance performance but also significantly reduce energy consumption and costs. As cloud computing and AI technology continue to evolve, further optimizing these workflows will be a crucial area of research.