Medical imaging instruments such as MRI scanners are complex dynamic systems whose non-ideal electrical and physical behavior significantly impacts image quality. This Research Internship focuses on developing models of MRI system behavior through calibration experiments, physical modeling, and dataādriven methods, with the goal of improving…
We are building new systems to optimize the millions of server nodes underlying the Microsoft Azure cloud. As a Research Intern you will be part of a dynamic and collaborative team chartered to understand and improve how hardware and software ingredients come together to form…
Small foundation models for the electric grid GridFM is a Microsoft Research initiative to build a foundation model (FM) for electric power grids, applying modern AI methodsāsimilar to large language/weather modelsāto complex grid physics. Traditional powerāflow solvers (like ACāOPF) are accurate but extremely slow, taking…
Geospatial Foundation Models ā deep learning architectures trained on large-scale Earth observation and remote sensing data ā have proved useful for object detection, semantic segmentation, and other prediction tasks on geospatial and geographic data. However, experts in domains most likely to benefit from applying these…
At Microsoft Research AI for Science, we believe deep learning has the potential to transform scientific modelling and discovery crucial for solving the most pressing problems facing society, including sustainable materials and discovery of new drugs. For our labs in Amsterdam (NL), Cambridge (UK) and…
WeāreāÆlooking for data scientists to help build the next generation of post-training methods for frontier models at Microsoft AI.āÆYouāllāÆjoin a small, high-impact team working across all stages of post-training, with a focus on evaluation design, high-quality training data, and scalable data pipelinesāÆforāÆstate-of-the-artāÆfoundation models. In this…
In enterprise search, building high-quality datasets at scale remains a central challenge due to the difficulty of acquiring labeled data. To resolve this challenge, we propose an efficient approach to fine-tune small language models (SLMs) for accurate relevance labeling, enabling high-throughput, domain-specific labeling comparable or…
We provide background on emerging challenges and future directions with media integrity and authentication methods, focusing on distinguishing AI-generated media from authentic content captured by cameras and microphones. We evaluate several approaches, including provenance, watermarking, and fingerprinting. After defining each method, we analyze three representative…
Modern generative and multimodal models increasingly rely on compact latent representations that trade and balance semantic richness with high-fidelity reconstruction. We introduce SALAD-VAE, a continuous and highly compact semantic Audio Variational Autoencoder, which operates in the frequency domain and achieves state-of-the-art compression with very low…
Robot Foundation Models (RFMs), also referred to as Vision-Language Action models (VLAs), have been attracting the attention of researchers and practitioners with a promise of generalizing robot behaviors across tasks, objects, and environments. The community has extensively studied RFMs' generalization capabilities in the vision and…
We study the task of determinant maximization under partition constraint, in the context of large data sets. Given a point set VāR^d that is partitioned into s groups V_1,...,V_s, and integers k_1,...,k_s where k=ā_i k_i, the goal is to pick ki points from group i…
For the past two decades, the DB community has devoted substantial research to take advantage of cheap clusters of machines for distributed data analytics -- we believe that we are at the beginning of a paradigm shift. The scaling laws and popularity of AI models…
