GenAI for Industry - AI-generated image compilation of industrial blueprints from renewable energy to transporation and manufacturing
Research for Industry

GenAI for Industry

How can industry benefit from generative AI?

Large language models (LLMs) are powerful tools that can generate natural language texts and answer complex questions across various domains. Microsoft Research is exploring the use of LLMs for various industry, aiming to solve domain specific challenges. Our work, highlighted in recent projects and publications, showcases applications to demonstrate how LLMs can transform vertical industries, like agriculture.

Our research is focused in these four areas of generative AI for industry:

  • Customizing LLMs
  • Small language models and edge
  • Multi-modal GenAI
  • Foundation models for industry
Satellite map of a town

Customizing LLMs

Scenario: A company wants to use a customized LLM that was built using their own documents, custom data. This requires customer data processing pipelines for finetuning or Retrieval Augmented Generation (RAG). The company also wants to use plugins that can integrate the LLM with their existing system of record and provide interactive feedback to the internal or external customers. The generated answers from complies with the privacy and ethical standards of the industry and company policies.

Small language models and edge

Scenario: A company wants to use a small LLM to create immersive and dynamic narratives for their online versions. The LLM needs to run on the edge devices without relying on constant cloud servers or internet connection. The LLM also needs to adapt to the users’ preferences and actions and generate realistic and diverse dialogues and scenarios. The company also wants to use a plugin that can optimize the LLM for low-latency and mobile operation, as well as ensure safety and fairness.

Multi-modal GenAI

Scenario: A company wants to use a multi-modal LLM to generate captions and summaries for their video and audio content. The LLM needs to be able to process and understand different types of data, such as images, speech, music, and other non-text formats. The LLM also needs to be able to generate coherent and informative captions and summaries that capture the main points and emotions of the content. The company also wants to use a plugin that can enhance the quality and diversity of the LLM outputs, as well as provide feedback and editing options to the content creators.

Foundation models for industry

Scenario: A company wants to use simulations and custom computation to build a custom AI model that will answer target scenario questions. These are foundational models, custom trained for specific domains. These along with previous models can help holistically answer customers’ queries.

Key industries

FarmVibes - man walking through a wheat field towards a distant barn (Photo by Dan DeLong for Microsoft)

Agriculture

Answers can help address the specific queries of different personas within the Agri-Food ecosystem, from farmers and policymakers, consumer, retailers to financial service providers and sustainability consultants.

engineer check and control automation robot arms machine in intelligent factory industrial, welding robotics and digital manufacturing operation

Manufacturing

Streamlines operations, enhancing efficiency and innovation across the manufacturing sector, from production floor managers to supply chain analysts.

Image of a lady shopping in a grocery store

Consumer goods / Retail

Transforms the retail experience, offering personalized assistance to store managers, inventory specialists, and customer service representatives. Optimizes product lifecycle management and market analysis, benefiting brand managers and consumer insights analysts in the consumer goods industry.

sustainability: photo of solar panels at sunrise

Sustainability

Supports sustainability initiatives, providing actionable insights for environmental consultants and corporate sustainability officers to drive eco-friendly practices in their organizations.