cv
Basics
| Name | Yu Gu (顾禹) |
| Label | Research/Applied Scientist |
| aidengu001@gmail.com | |
| Phone | +1-858-342-9893 |
| Url | https://www.linkedin.com/in/YourLinkedInProfile |
| Summary | I’m Yu Gu (顾禹), also known as “Aiden”— currently at Microsoft Research and former co-founder of a Series C startup. I specialize in large-scale foundation models, multimodal learning, and enterprise AI deployment. My work includes building PubMedBERT (20M+ downloads) and other high-impact AI systems showcased by Forbes, CNBC, and the World Economic Forum. I’m passionate about transforming advanced research into tangible solutions that shape the future of AI. |
Work
-
2020.01 - Present Senior Applied Scientist
Microsoft
Specializing in building domain-specific foundation models and delivering enterprise AI solutions.
- Developed PubMedBERT, one of the first domain-adaptive LLMs (20M+ downloads, 2000+ citations, ACM HLTH Best Paper of the Year)
- Created BiomedParse (Nature Methods) and BiomedJourney for multimodal imaging and text-to-image generation in healthcare
- Deployed multi-agent AI systems for tumor board decision-making, showcased at the World Economic Forum
- Led cross-functional teams (scientists, engineers, PMs, designers) to build enterprise-grade, cloud-deployed AI solutions
- Optimized LLM inference with retrieval-based methods (FAISS), improving real-time search efficiency and scalability
-
2016.06 - 2020.01 Machine Learning Scientist / Co-founder
Med Data Quest Inc.
Co-founded an AI startup that reached Series C acquisition, driving the transition from research prototypes to enterprise deployments.
- Led a 10+ person international team designing hierarchical AI pipelines for medical data processing
- Developed full-stack NLP solutions, securing top-5 rankings in N2C2 Challenges (2016, 2018, 2019)
- Oversaw product strategy and R&D, bridging cutting-edge research with clinical and enterprise needs
Education
-
Beijing, China
Publications
-
2022.01.01 BiomedJourney: Temporal AI for Patient Outcome Simulation
arXiv
Text-to-image generation model enabling realistic medical image synthesis and dynamic disease progression simulation.
-
2022.01.01 PubMedBERT: Domain-Specific Language Model Pretraining
ACM Health
Introduced a domain-adaptive LLM addressing tokenization and terminology gaps where general-purpose models fail.
-
2021.10.01 GigaPath: A Whole-Slide Foundation Model
Nature
A large-scale model tailored for pathology data, supporting high-resolution slide analysis.
-
2021.06.01 BiomedParse: Image Parsing for Everything, Everywhere
Nature Methods
Universal segmentation foundation model for multimodal imaging (CT, MRI, pathology).
-
2021.01.01 UniversalNER: LLM Distillation for Open NER
ICLR
Scalable named entity recognition framework leveraging LLM distillation and retrieval-enhanced learning.
Skills
| Machine Learning & AI | |
| Large-Scale Foundation Models | |
| Multimodal Learning | |
| Enterprise AI Deployment | |
| Retrieval-Augmented Generation (RAG) | |
| Multi-Agent AI Systems | |
| Cloud-Based Infrastructure | |
| Model Distillation | |
| Generative AI |
Interests
| AI Research & Deployment | |
| Multimodal Applications | |
| Foundation Model Adaptation | |
| Scalable Inference | |
| Model Interpretability | |
| Agentic Modeling |