Microsoft Research (MSR) is the research division of Microsoft Corporation, established in 1991. It operates numerous laboratories worldwide, focusing on advancing the state of the art in computer science, software engineering, and related interdisciplinary fields. MSR is known for its high-impact academic publications, open-source contributions, and collaborative partnerships with universities and research institutions.
1 History
1.1 Origins and founding (1991–2000)
Microsoft Research was founded in 1991 by Nathan Myhrvold, then Microsoft’s chief technology officer, with the goal of creating a basic research arm similar to corporate labs like Xerox PARC and Bell Labs. The initial lab was located in Redmond, Washington, and recruited leading computer scientists. Early hires included Richard Rashid, who became the first director, and researchers in programming languages, operating systems, and graphics. During the 1990s, MSR established the foundations for later breakthroughs, such as the Tablet PC concept and early work on natural language processing.
1.2 Expansion era (2001–2010)
In the early 2000s, MSR expanded globally, opening labs in Cambridge, UK (1997–2001 full establishment), Beijing, China (2005), and Bangalore, India (2005). The division grew to hundreds of researchers, diversifying into areas like machine learning, computational biology, and distributed systems. Major projects included the development of the .NET Framework’s language innovations and the precursor to the Kinect motion-sensing system (Project Natal). This era also saw increased collaboration with academia through joint research centers.
1.3 Modern developments (2011–present)
From 2011 onward, MSR intensified its focus on artificial intelligence, cloud computing, and open-source software. The lab contributed to Microsoft’s Azure AI platform and the development of the deep learning framework CNTK (Computational Network Toolkit). In 2016, MSR launched the Microsoft Research AI (MSR AI) initiative, consolidating AI-related efforts. The division also embraced open source, releasing projects such as the .NET runtime and Visual Studio Code under permissive licenses. Recent work encompasses responsible AI, quantum computing (through Station Q), and mixed reality research.
2 Research areas
2.1 Artificial intelligence and machine learning
2.1.1 Deep learning and neural networks
MSR has been a pioneer in deep learning, developing innovations such as the Deep Residual Network (ResNet) architecture, which achieved state-of-the-art results in image recognition. Researchers also contributed to the development of efficient training algorithms and the CNTK framework. Work on reinforcement learning led to systems like Project Malmo, which uses Minecraft as an AI training environment.
2.1.2 Natural language processing
Natural language processing (NLP) research at MSR includes foundational work on statistical machine translation, language models, and dialogue systems. The lab developed the Turing Natural Language Generation (T-NLG) model, one of the largest transformer-based models at the time of release. MSR also contributed to the creation of the Microsoft Language Understanding Intelligent Service (LUIS) and the popular open-source toolkit for sequence labeling, SeqLabel.
2.2 Systems and networking
2.2.1 Distributed systems
MSR has advanced distributed systems research through projects like the Orleans virtual actor framework, which simplifies building scalable cloud applications. The lab also contributed to the development of the Dryad and DryadLINQ data-parallel processing platforms, influencing later systems like Apache Spark. Research on distributed consensus protocols, such as Paxos variants, has informed Azure’s storage infrastructure.
2.2.2 Cloud computing
Cloud computing research at MSR includes work on resource management, serverless computing, and data center networking. The lab developed the Cluster Scheduler for Azure (e.g., Autopilot) and contributed to the Open Cloud Initiative. Research on energy-efficient data centers, including the use of underwater data centers (Project Natick), has been widely recognized.
2.3 Human-computer interaction
2.3.1 Accessibility technologies
MSR’s accessibility research focuses on tools for people with disabilities. Projects include Seeing AI, a mobile app that narrates the visual world for the blind; the Soundscape system for spatial audio navigation; and Captions AI for real-time speech-to-text. The lab also developed the Eyegaze communication system for individuals with motor impairments.
2.3.2 Mixed reality
Mixed reality research underpins Microsoft’s HoloLens and Windows Mixed Reality platforms. MSR contributed to advanced computer vision algorithms for spatial mapping, gesture recognition, and photorealistic 3D rendering. The lab’s work on light-field displays and computational photography has directly influenced the design of head-mounted displays.
2.4 Theoretical computer science
2.4.1 Algorithms and complexity
MSR’s theory group has made contributions to approximation algorithms, online algorithms, and algorithmic game theory. Researchers have published influential work on the hardness of approximation, the design of efficient data structures, and the theory of spectral graph theory. These results have applications in network routing, search, and recommendation systems.
2.4.2 Cryptography
Cryptography research at MSR includes advancements in lattice-based cryptography, homomorphic encryption, and secure multiparty computation. The lab developed the Microsoft SEAL library for homomorphic encryption, now widely used in privacy-preserving machine learning. Work on verifiable computation and zero-knowledge proofs has contributed to blockchain and digital identity systems.
2.5 Health and life sciences
2.5.1 Computational biology
MSR’s computational biology group applies machine learning and data science to genomic and proteomic analysis. Projects include the Microsoft Genomics service, which leverages cloud computing for processing DNA sequences. Researchers have also developed algorithms for predicting protein folding (e.g., in collaboration with the AI for Health initiative) and for analyzing microbiome data.
2.5.2 Medical imaging
Medical imaging research at MSR covers image segmentation, disease detection, and clinical decision support. The lab developed InnerEye, an open-source tool for automated segmentation of tumors in radiology scans. Work on deep learning for chest X-ray analysis and pathology slide interpretation has been published in top medical journals and integrated into Microsoft’s healthcare solutions.
3 Notable projects and products
3.1 Research prototypes
3.1.1 Project Adam
Project Adam was a large-scale deep learning system developed at MSR in 2014. It demonstrated state-of-the-art image classification using a distributed, asynchronous training approach on thousands of machines. The system was capable of recognizing over 22,000 object categories and influenced later distributed training frameworks.
3.1.2 Coco (Conversational AI)
Coco (originally named “Cortana Conversations”) is a research prototype for building robust conversational AI agents. It focuses on goal-directed dialogue, multi-turn reasoning, and handling of ambiguous queries. Coco’s architecture has informed the development of Microsoft’s Power Virtual Agents and Azure Bot Service.
3.2 Open-source contributions
3.2.1 .NET Core
MSR played a key role in the open-sourcing of .NET Core in 2014, releasing the runtime, libraries, and compilers under the MIT license. This move transformed .NET into a cross-platform, community-driven framework. Many performance and language improvements originated from MSR’s research on just-in-time compilation (e.g., the RyuJIT compiler).
3.2.2 VSCode (in collaboration)
Visual Studio Code, a popular code editor, was developed by Microsoft in collaboration with MSR researchers. The editor’s architecture, including the Language Server Protocol and the Monaco editor engine, drew on MSR’s work in program analysis, code intelligence, and user interface design. VSCode is now open source and widely adopted.
3.3 Technologies transferred to Microsoft products
3.3.1 Bing search algorithms
Several core algorithms for Bing’s search engine originated from MSR. These include the ranking algorithm RankNet (a neural network-based learning-to-rank model), the click-through prediction system, and the entity-relation indexing approach. MSR also contributed to the development of Bing’s image and video search capabilities.
3.3.2 Azure AI services
MSR’s research directly underpins many Azure AI services, such as Azure Cognitive Services (vision, speech, language decision) and Azure Machine Learning. Specific transfers include the speech recognition models used in Azure Speech-to-Text, the language models powering Azure Language Understanding (LUIS), and the computer vision algorithms in Azure Computer Vision.
4 Laboratories and locations
4.1 Main labs
4.1.1 Redmond (headquarters)
The Redmond lab, located on Microsoft’s main campus in Washington, is the oldest and largest MSR facility. It houses research groups in AI, systems, theory, and human-computer interaction. The lab also hosts the Microsoft Research AI center and the Station Q quantum computing group.
4.1.2 Cambridge (UK)
The Cambridge lab (often referred to as MSR Cambridge) was established in 1997 and formally opened in 2001. It focuses on machine learning, computational science, and security. Notable achievements include the development of the C# programming language’s early design and the F# language. The lab has strong ties with the University of Cambridge.
4.2 Regional labs
4.2.1 Beijing (Asia)
The Beijing lab, established in 2005, is MSR’s largest research lab in Asia. Its research areas include natural language processing, computer vision, and software analytics. The lab has produced influential work on Chinese language processing and contributed to Microsoft’s translation services. It also runs an internship program that attracts top Chinese university students.
4.2.2 Bangalore (India)
The Bangalore lab, also opened in 2005, focuses on software engineering, systems research, and AI for emerging markets. Research includes work on multilingual NLP for Indian languages, low-cost sensor networks, and mobile computing. The lab collaborates closely with Indian institutes like the Indian Institutes of Technology (IITs).
5 Collaborations and academic outreach
5.1 Joint research centers
5.1.1 Microsoft Research–MIT Lab
The Microsoft Research–MIT Laboratory for Social Machines, established in 2014, was a joint venture to study social media data and human communication. The lab applied machine learning to analyze public discourse, identify misinformation patterns, and develop tools for journalists. It operated until 2019.
5.1.2 Microsoft Research–INRIA Centre
The Microsoft Research–INRIA Joint Centre, founded in 2006 in France, focused on computer science fundamentals. Research areas included programming languages, formal verification, and bioinformatics. The centre produced significant contributions to the Coq proof assistant and the F* programming language. It concluded in 2019.
5.2 Fellowship and internship programs
5.2.1 PhD Fellowship Program
The Microsoft Research PhD Fellowship Program awards financial support and mentoring to outstanding doctoral students worldwide. Fellowships are offered in multiple regions (e.g., North America, Europe, Asia) and cover tuition, stipend, and travel expenses. Recipients are selected based on research potential in fields aligned with MSR’s interests.
5.2.2 Research Internships
MSR’s research internship program attracts graduate and undergraduate students from around the globe. Interns work on cutting-edge research projects under the supervision of MSR scientists, often resulting in peer-reviewed publications. The program is highly competitive, with a typical duration of 12 to 16 weeks.
6 Impact and recognition
6.1 Awards and honors
6.1.1 Turing Award recipients
Several MSR researchers and collaborators have received the ACM Turing Award, the highest honor in computer science. Notable recipients include Leslie Lamport (2013) for contributions to distributed systems; Butler Lampson (1992) for personal computing; and Charles Thacker (2009) for early computer design. MSR also employed or hosted winners such as John McCarthy (though not full-time).
6.1.2 ACM Fellow members
Many MSR researchers are ACM Fellows, recognized for their contributions to computing. Examples include Cynthia Dwork for differential privacy, James Larus for programming languages, and Kate Crawford for ethical AI research. As of 2024, over 100 current and former MSR employees hold this distinction.
6.2 Influence on industry and society
6.2.1 Open-source ecosystem
MSR has significantly shaped the open-source landscape through contributions like the .NET runtime, Visual Studio Code, the TypeScript language, and the CNTK and Quantum Development Kit. These projects have millions of users and have fostered large developer communities. MSR also supports open-source foundations, such as the .NET Foundation and the Open Source Initiative.
6.2.2 Policy and ethics research
MSR engages in research on the societal implications of technology, including fairness, accountability, transparency, and privacy. The lab’s Ethics & Society group publishes guidelines and toolkits for responsible AI development. Work on differential privacy (by Cynthia Dwork and colleagues) has influenced global data protection regulations, including those in healthcare and census processes.