Research
Discover an index of datasets, SDKs, APIs and open-source tools developed by Microsoft researchers and shared with the global academic community below. These experimental technologies—available through Azure AI Foundry Labs (opens in new tab)—offer a glimpse into the future of AI innovation.
A framework/SDK for program synthesis from input-output examples, with pre-built applications for data wrangling, Jupyter integration for data scientists, repetitive code editing, text manipulations, and extraction from webpages. Started at MSR, now developed by a…
Vowpal Wabbit is a machine learning system which pushes the frontier of machine learning with techniques such as online, hashing, allreduce, reductions, learning2search, active, and interactive learning. There is a specific focus on reinforcement learning…
Vale (Verified Assembly Language for Everest) is a tool for constructing formally verified high-performance assembly language code, with an emphasis on cryptographic code. It uses existing verification frameworks, such as Dafny and F*, for formal…
EverCrypt (opens in new tab) is a high-performance, cross-platform, formally verified modern cryptographic provider distributed as a combined C/ASM library. EverCrypt packages cryptographic implementations from the HACL* and ValeCrypt projects, and automatically picks the fastest…
GitHub Publication Publication Publication Publication Publication
Supporting code for the Model-Based Machine Learning book. This project contains the sample code and test data for the freely available online book on model-based machine learning published at http://mbmlbook.com/ (opens in new tab). The…
This data set contains accelerometer and gyroscope recordings from over 200 participants performing various gym exercises. This data set is described in more detail in the associated manuscript: Morris, D., Saponas, T. S., Guillory, A.,…
A portable volatile memory acquisition tool for Linux. AVML is an X86_64 userland volatile memory acquisition tool written in Rust, intended to be deployed as a static binary. AVML can be used to acquire memory…
AMDIM (Augmented Multiscale Deep InfoMax) is an approach to self-supervised representation learning based on maximizing mutual information between features extracted from multiple views of a shared context.
MazeExplorer is a customisable 3D benchmark for assessing generalisation in Reinforcement Learning. It is based on the 3D first-person game Doom and the open-source environment VizDoom. This repository contains the code for the MazeExplorer Gym…