UniRG: Scaling medical imaging report generation with multimodal reinforcement learning
AI can help generate medical image reports, but today’s models struggle with varying reporting schemes. Learn how UniRG uses reinforcement learning to boost performance of medical vision-language models.
AI for Low-Resource Languages
< AI for Good Lab Language is more than communication—it sustains culture, identity, and opportunity. Yet most of the world’s languages remain underrepresented in today’s AI systems because they have limited digital content, few high-quality…
Microsoft Research India 2025 Highlights
Here are some highlights of the research and impact at Microsoft Research India in 2025, along with some glimpses of the lab milieu.
Research Intern – Microsoft Teams (PhD)
This role is specifically for a Microsoft Teams Research Intern — you’ll work on advanced research that supports innovation in collaboration and communication technologies. Projects may involve exploring cutting-edge areas such as AI-driven productivity tools,…
Research Intern – AIP AI Knowledge Multimodal AI
The mission of the AI Platform team is to advance the state of the art in AI and create breakthroughs that power next-generation applications. As a Research Intern, you’ll contribute to cutting-edge research in multimodal…
Research Intern – Office of Applied Research
The Office of Applied Research brings together experts in Machine Learning (ML) (including reinforcement learning and synthetic data generation), Large Language Models (LLM) (development, evaluation, and fine-tuning), and Human-Centered AI (such as collaboration intelligence and…
Leveraging Loanword Constraints for Improving Machine Translation in Low-resource Settings
Translating from high-resource to low-resource languages like Emakhuwa remains a challenge due to limited parallel data, orthographic variation, and frequent loanwords and code-switching. In this talk Felermino will discuss how to apply lexicon-guided neural machine…
NeurIPS 2025
Microsoft is pleased to have over 150 accepted papers at NeurIPS 2025, scheduled for December 2–7, 2025, in San Diego, CA, USA.