The eIQ Time Series Studio features an automated machine learning workflow that streamlines the development and deployment of time series-based machine learning models across MCU-class devices, such as the MCX portfolio of MCUs or the i.MX RT portfolio of crossover MCUs.
The GenAI Flow provides the building blocks for Large Language Models (LLMs) that power generative AI solutions. Designed to be used with MPUs such as NXP’s i.MX family of applications processors, these generative AI solutions make it easier to deploy intelligence at the edge by training LLMs on specific contextual data.
For example, an appliance equipped with an LLM trained on the user manual could converse with a user in natural language about how to access certain features, perform certain tasks or otherwise optimise usage and maintenance.
Deploying AI at the edge offers numerous benefits, including reduced latency, improved user privacy, and reduced energy consumption and NXP’s expansion of the eIQ Toolkit makes those deployments significantly easier and faster, giving developers access to a wider range of model types - from generative AI to time series-based models to vision-based models. Users are also able to deploy these models on a wider range of edge processors.
“AI is the key to a world that anticipates and automates based on user wants and needs, but it must be developed in a way that is practical for edge deployment,” said Charles Dachs, Senior Vice President and General Manager, Industrial and IoT, NXP Semiconductors. “With ready-to-use tools suitable for both small AI models on MCUs like the MCX portfolio, crossover MCUs like the i.MX RT700 as well as larger, generative AI models running on more powerful devices like the i.MX 95 applications processor, NXP is delivering a breadth of options for developers across the full spectrum of AI models and AI-enabled edge processors. NXP is making edge AI truly practical for developers across a wide range of markets.”
The eIQ Time Series Studio simplifies and reduces time required for development and deployment of time series-based AI models. It supports a wide range of input signals, including voltage, current, temperature, vibration, pressure, sound, time of flight, among others, as well as combinations of these for multi-modal sensor fusion.
The automatic machine learning capability enables developers to extract meaningful insights from raw time-sequential data and quickly build AI models tailored to meet performance, memory, flash storage size, and accuracy criteria.
The tool offers a comprehensive development environment, including data curation, visualization and analysis, as well as model auto-generation, optimisation, emulation, and deployment. Its intuitive interface allows software developers to create optimised anomaly detection, classification and regression libraries, all without the need for deep data science or AI expertise.
NXP’s GenAI Flow makes generative AI applications accessible on edge devices. The software pipeline provides a means to optimise generative models like LLMs.
The GenAI Flow also provides Retrieval Augmented Generation (RAG), a method to securely fine-tune models on domain-specific knowledge and private data without exposing sensitive information to the model or processor providers.
By chaining together multiple modules in a single flow, customers can customise LLMs to their task and optimise them for deployment at the edge.
