Qiran Jia
Machine Learning Scientist · Statistician · Ph.D. Candidate, USC
Probabilistic & deep learning · Representation learning · Multimodal data · Computational biology
About
I'm a machine learning scientist and Ph.D. candidate in Biostatistics at the University of Southern California, finishing in January 2027. I build statistical, variational Bayes, and deep-learning methods, along with the software that puts them into practice, to turn high-dimensional, multimodal data into reliable discoveries, causal insights, and predictions. Most of my work applies them to biomedical data: bulk and single-cell multi-omics, longitudinal live-cell imaging, and 3D medical imaging.
Recent work includes Praxis-BGM, a semi-supervised Bayesian transfer-learning method for Gaussian mixture models via natural-gradient variational inference, implemented in JAX (Bioinformatics, 2026), and LUCID, a quasi-mediation framework for integrative multi-omics clustering and inference with high-dimensional exposures and outcomes, released as an R package with 36,000+ downloads (R Journal, 2024). In summer 2026 I interned on the computational biology team at Cellanome, where I built a GPU-accelerated deep-learning trajectory autoencoder that learns cell embeddings from live-cell imaging sequences (60× faster than statistical baselines and adopted in internal software), then fused them with transcriptomics to reveal cell states missed by snapshot morphology alone. Earlier, at NYU, I built a CNN–vision-transformer for 3D MRI segmentation and co-developed DeepFDR (AISTATS, 2024), deep-learning-based false discovery rate control for 3D neuroimaging.
My doctoral research is advised by David V. Conti (Conti Lab, now chair of Biostatistics & Informatics at the Colorado School of Public Health), and my master's work was advised by Hai Shu at NYU.
I'm most excited by probabilistic and deep learning that makes complex, multimodal data interpretable and actionable, whether that means single-cell and multi-omics data or other large-scale scientific and real-world data, and by pairing rigorous inference with models that ship.
Details on the experience and publications pages.
Interests
- Representation learning
- Deep learning & computer vision
- Bayesian & variational inference
- Transfer learning
- Causal inference
- Statistical inference & multiple testing
- Multimodal data integration
- Single-cell & multi-omics
- Medical & cell imaging