锦鲤求职岗位描述
岗位职责
Introduction:
Our group builds multimodal foundation models that integrate spatial multi-omics with computational pathology. Spatial multi-omics here covers spatial transcriptomics, proteomics, metabolomics and epigenomics. Computational pathology covers H&E and IHC/mIF whole-slide images. Together, these models capture both the tissue morphology and the molecular landscape of disease at cellular and sub-cellular resolution.
This sub-direction focuses on turning these foundation models into real-world clinical and translational research tools. We work closely with clinicians, pathologists and biomedical researchers to find high-value research needs. We turn those needs into well-defined computational tasks. Then we work with the algorithm team to develop domain-specific vertical models and AI agents. Representative application scenarios include:
Tumor microenvironment (TME) characterization: spatial cell-type deconvolution, cell–cell interaction and niche discovery, and tertiary lymphoid structure (TLS) identification.
Virtual spatial omics: inferring spatial gene and protein expression from routine H&E slides to expand clinical and research scenarios.
Biomarker discovery and clinical prediction: molecular subtyping, prognosis stratification, and prediction of response to immunotherapy, targeted therapy and chemotherapy.
Research and diagnostic AI agents: LLM-driven agents that orchestrate pathology and omics models, knowledge bases and analysis tools. These agents support clinician-scientists in hypothesis generation, cohort analysis and report interpretation.
Students in this sub-direction work across the full translational pipeline. That pipeline runs from clinical question definition, multi-center data curation and model adaptation (fine-tuning, prompting, agent workflow design) to rigorous benchmarking, clinical validation, and publication or deployment. This position suits students who want to connect cutting-edge AI with meaningful biomedical impact.
Affiliated Branch(es):
Southwest Branch (Chongqing)
岗位要求
(1)Preferred Background:
- A degree (Bachelor's/Master's) in Bioinformatics, Computational Biology, Biomedical Engineering, Computer Science, Artificial Intelligence, Biostatistics, or a related field.
- Applicants from Clinical medicine, Pathology, Oncology or Life Sciences who have strong computational training are also strongly encouraged to apply.
- Familiarity with at least one of the following is a plus: spatial omics technologies (e.g., 10x Visium/Visium HD, Xenium, CosMx, MERFISH, Stereo-seq, CODEX/IMC), single-cell omics, or digital pathology.
(2)Core Technical Skills:
- Proficiency in Python. Experience with PyTorch is expected, and R is a plus.
- Hands-on experience with omics analysis tools such as Scanpy, Seurat, Squidpy or AnnData, and/or pathology image tools such as OpenSlide, QuPath or CLAM.
- Ability to process large-scale, heterogeneous biomedical data: whole-slide images, sparse high-dimensional matrices, and clinical tables.
- Basic statistics and survival analysis (e.g., Cox models, Kaplan–Meier, ROC/AUC, multiple-testing correction).
- Working knowledge of Linux, Git and GPU/HPC computing environments.
(3)Modeling Competency:
- Solid understanding of deep learning for vision and sequences: CNNs, Vision Transformers, self-supervised and contrastive learning.
- Experience with or strong interest in multimodal learning, such as cross-modal alignment (e.g., CLIP-style), fusion strategies, and multiple-instance learning for WSIs.
- Ability to adapt pretrained or foundation models to downstream tasks through linear probing, fine-tuning, or parameter-efficient methods (LoRA, adapters).
- Familiarity with LLMs and agent frameworks (e.g., tool calling, RAG, LangChain/LangGraph, MCP) is highly valued for agent-oriented projects.
(4)Preferred Qualifications:
- First-author or co-author publications, preprints, or conference papers in computational pathology, spatial omics, medical AI or related areas.
- Experience in clinical or translational research projects, such as collaborating with hospitals, analyzing patient cohorts, or developing biomarkers.
- Contributions to open-source tools, Kaggle/MICCAI-type challenges, or public benchmarks.
- Experience with pathology foundation models (e.g., UNI, CONCH, Virchow, Prov-GigaPath) or spatial/single-cell foundation models.
- Good written and spoken English for scientific communication. Knowledge of Chinese is a plus but not required.
(5)Personal Traits:
- Translational mindset: genuinely motivated to solve real clinical problems, not only to improve benchmark metrics.
- Interdisciplinary communication: able to "translate" between clinicians, biologists and algorithm engineers, and comfortable working across team boundaries.
- Self-driven and problem-oriented: able to break an ambiguous research need into concrete, executable tasks, and to work independently.
- Rigor and integrity: careful handling of patient data, reproducible work, and honest reporting of results.
Team spirit: collaborative, responsible, and willing to share knowledge within a multidisciplinary team.