Explore more of Ryen White’s work at Microsoft Research
Learn more
Research has always been an integral part of Microsoft’s identity, driving our role as a global technology leader. Since 1991, Microsoft Research has dedicated itself to a fundamental research approach—advancing knowledge, deepening our understanding of the world, and exploring how technology can empower people and organizations. Through its extensive body of publications and a commitment to openly sharing its work, Microsoft Research continues to collaborate with the global research community to drive breakthroughs in AI and beyond. Together, we are pushing the boundaries of what’s possible to extend human capability, create value for our customers, and deliver broad societal benefit.
Transforming research with AI at Microsoft Research
With Microsoft leading the charge in AI, Microsoft Research teams are at the forefront, using our expertise, fostering thought leadership, and driving innovation in AI and research. AI is at the center of many of Microsoft Research’s pioneering projects, from helping researchers analyze massive datasets within seconds, inventing new AI solutions that would benefit humanity, and collaborating with the global research community through the Accelerating Foundation Models Research (AFMR) program.
Microsoft Research is in a unique position where it can not only adopt and use AI but also invent AI. We have made significant investments in AI, building new AI methods, models, and technologies. To infuse AI more deeply into the research process, an experimental initiative is underway, helping teams move faster, think bigger, and share more effectively. This initiative breaks down into three key strategies: using, infusing, and diffusing AI across the organization.
Using AI (Tools and Operations) focuses on optimizing access and advancement of AI.
Infusing AI (Research and Development) is about maximizing the potential of AI to revolutionize research processes.
Diffusing AI (Connectivity and Information Flow) ensures rapid sharing of AI insights, tools, and learnings with others, both inside and outside of Microsoft Research.
The goal is not just to adopt AI, but to augment and reinvent the way research is done—empowering everyone in the organization to achieve more.
The integration of AI into research processes at Microsoft Research provides valuable insights for researchers and businesses. Using AI can accelerate innovation cycles, improve operational efficiency, and lead to the development of cutting-edge tools and products. These advancements highlight how AI can reinvent traditional workflows, streamline operations, and drive growth and profitability, making it a strategic focus for organizations to implement.
GraphRAG: Advancing research with knowledge graphs
GraphRAG is a modular graph-based retrieval augmented generation (RAG) system that uses large language models to create knowledge graphs from raw text. This technique enhances large language model performance on private datasets by providing structured data and summaries, making it easier for researchers to extract meaningful insights from complex data.
The changes that are happening in AI right now, they really are surprising. The capabilities are expanding so quickly. I think of it as kind of an accelerator. Everything that we do in research, we can do faster, we can ask more questions, and this has all been kind of a warp speed thing.
—Nathan Evans, Principal Software Architect at Microsoft Research
Data Formulator: Transforming data into insights
Exploring how AI can help analysts create rich data visualizations
Read the blog
Data Formulator is an innovative tool designed to help researchers quickly explore and analyze data. By using AI, Data Formulator lets users to create rich visualizations without the need for extensive programming knowledge. This tool combines AI and interactive approaches to communicate visualization intent, making data analysis more accessible and efficient
AI really speeds up our experimentation process. In the past, we really needed to do a lot of hacking over weeks to experiment on designs. But now we can have a high-level thought, we can do the prototype in a short amount of time, and we can start thinking on top of that.
—Chenglong Wang, Senior Researcher at Microsoft Research
Accelerating Foundation Models Research: Democratizing AI research
The Accelerating Foundation Models Research (AFMR) program provides academic researchers with access to state-of-the-art foundation models hosted on Microsoft Azure through Microsoft Azure AI services. This initiative fosters a global AI research network and offers robust, trustworthy models that help further research in disciplines ranging from scientific discovery and education to healthcare, multicultural empowerment, legal work, and design.
The AFMR program works with the broader academic research community to explore different aspects of foundation models to accomplish three goals:
Goal 1: Align AI with shared human goals, values, and preferences
This involves enhancing the safety, robustness, sustainability, responsibility, and transparency of AI models. One notable project aligned to this goal is “ERBench: An Entity-Relationship based Automatically Verifiable Hallucination Benchmark for Large Language Models.” For this project, researchers created ERBench which helps in evaluating and improving the accuracy and reliability of AI-generated content. This ensures that AI models align with human values and reduces the risk of misinformation.
Goal 2: Improve AI-human interactions
The second goal focuses on improving AI-human interactions by increasing trust, human ingenuity, creativity, and productivity while reducing the risk of developing AI that is harmful for individuals and society. The project “As Generative Models Improve, People Adapt their Prompts” explores how prompting changes as generative AI models improve. Results showed that participants using more advanced models produced better, longer, and more descriptive prompts. This research provides valuable insights into the evolving dynamics between humans and AI, helping to create more intuitive and effective AI systems.
Goal 3: Accelerate scientific discovery
The third goal is to accelerate scientific discovery through proactive knowledge discovery, hypothesis generation, and multimodal data generation. One project that exemplifies this goal explored “Artificial Intelligence–Based Copilots to Generate Causal Evidence.” In this initiative, large language models were explored as causal “copilots” to help identify flaws in medical study designs. These models could assist researchers by providing expert guidance in study design, improving the accuracy of conclusions drawn from the studies.
AI is really important in research because AI has the potential, the huge potential to really accelerate the research, which is needed to address some of the greatest challenges of today and tomorrow.
—Evelyne Viegas, Technical Advisor at Microsoft Research
The next frontier: Looking ahead to the future of AI in research
As scientific research evolves in an era powered by AI and cloud technologies, the opportunities for innovation, collaboration, and global impact are unprecedented. From accelerating scientific discovery to improving human-agent alignment, foundation models are reshaping how research is conducted, shared, and scaled. Looking ahead, researchers and institutions must not only embrace these tools but also build robust frameworks for adoption, and evaluation.
There is still much more for us to explore on how we can advance research at Microsoft and we’re just getting started.
GIPHY App Key not set. Please check settings