PI · Neutrino AI Lab · IHEP-CAS
Scaling AI4S · research agents · HPC · particle physics
I build computational tools for neutrino physics, scientific machine learning, and AI-assisted research.
01 · RESEARCH
Four directions, one thesis — reliable intelligence operating at scale on real science.
DIRECTION 1 · SCALE
AI-for-Science in physics at production scale — pipelines on real detector data.
DIRECTION 2 · AGENTS
Agents that do research the way researchers do — on a gated, verification-first runtime.
DIRECTION 3 · HPC
Making reconstruction and statistical analysis fast — CPU/GPU performance, parallelization, and search-based tuning.
03 · PROJECTS
A terminal AI agent for complex research tasks. Multi-provider (DeepSeek, GLM, and Claude via the IHEP gateway) behind one OpenAI client, with fast/strong model slots, an async prompt queue, and built-in file & shell tools.
A transformer foundation model for the JUNO detector: self-supervised pre-training on per-PMT hit patterns, with fine-tuned heads for point-like energy/vertex reconstruction, muon-track reconstruction, and event classification.
An in-browser explorer for reactor ν̄ₑ oscillations — drag the mixing angles and Δm², slide Δm²₃₁ across zero to switch mass ordering, and watch both the survival probability and its Fourier power spectrum respond.
04 · NEWS & TALKS
The Borexino solar-neutrino program is awarded the EPS Giuseppe & Vanna Cocconi Prize.
Prospects of neutrino mass ordering and solar neutrinos with JUNO
2019.08.10–16plenary Lake Louise Winter Institute, Canada
Status and physics of JUNO
2018.08.12–18plenary NuFACT — 20th Int. Workshop on Neutrinos from Accelerators, Virginia Tech, USA
Borexino Phase-II results and prospects of CNO detection
2018.07.16–19plenary International Symposium of Neutrino Frontier, Quy Nhon, Vietnam