Chen Jin
Associate Principal Ai Scientis @阿斯利康
Signup · Get unlimited contacts
WORK HISTORY
Associate Principal Ai Scientis @阿斯利康
Cambridge, GB
Post-training multimodal models with enhanced vision-language alignment for healthcare challenges- Developed Lavender, a supervised fine-tuning method that boosts Llama-3.2-11B performance by up to 30% across 20 tasks ([Lavender Page](https://astrazeneca.github.io/vlm))- Created Multi-Concept Prompt Learning (MCPL), a mask-free prompt learning approach leveraging Stable Diffusion ([MCPL Page](https://astrazeneca.github.io/mcpl.github.io/))- Authored and co-authored impactful publications (ICML 2024, ECCV 2024 Oral, ICLR 2025, and under review) on image customization, local diffusion for hallucination reduction, and inference-time ensembling- Supervised multiple internship projects advancing research in grounded segmentation, causal discovery, and protein generation.
EDUCATION
Heriot-Watt University
Doctor of Philosophy - PhD, Digital Rock Analysis
SKILLS
ABOUT CHEN JIN
Jin leads frontier AI research in multimodal LLMs and agentic AI for healthcare and is the creator of Lavender—a post-training method that boosts Llama-3.2-11B with Stable Diffusion.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.