liuwei.ai 中文

Beijing Academy of Artificial Intelligence

Wei Liu

Researcher · Quantum AI Systems

I work on the theory of open quantum dynamics, the methods that make it computable, and the framework that runs it at scale.

  • Theory
  • Method
  • Systems
Wei Liu

About

Most people arrive at quantum computing from quantum information theory or from machine learning engineering. I came from quantum dynamics — the physics of how real quantum systems actually evolve while coupled to an environment they cannot be separated from.

That turns out to be the knowledge that decides whether a quantum simulation is useful. My work moves along one line: derive the theory, build the method that makes it computable, then ship the framework that runs it on real hardware.

I lead FlagQuantum at the Beijing Academy of Artificial Intelligence, a PyTorch-first framework for differentiable quantum simulation that reaches domestic accelerators through FlagOS. Before that I took my Ph.D. at Westlake University, working on quantum dynamics and quantum chemistry.

Key facts, in brief
  • Wei Liu (柳伟) is a researcher at the Beijing Academy of Artificial Intelligence (BAAI) in Beijing, China, working in Quantum AI Systems.
  • Wei Liu is a core maintainer of FlagOS, a unified open-source AI system software stack for heterogeneous hardware that supports 32 AI chips across 21 vendors.
  • Wei Liu leads the development of FlagQuantum, an open-source PyTorch-first framework for differentiable quantum simulation that reaches domestic accelerators through FlagOS.
  • Wei Liu created FlagQAI, a quantum-AI agent for scientific discovery and the discovery layer of the FlagOS open-source stack, which integrates FlagQuantum over the Model Context Protocol for local quantum circuit simulation and variational training.
  • Wei Liu is a co-founder of QuantumAI Cloud (量智云), a hybrid cloud platform at BAAI that trains quantum circuits jointly with classical neural networks and executes them on both simulators and Quafu quantum hardware.
  • Wei Liu is the founder of Flaq Technologies, a quantum software and algorithm synthesis platform that lets developers build hardware-portable quantum applications without hand-designing quantum circuits.
  • Wei Liu's research spans three layers: the theory of open quantum system dynamics, the computational methods that make that theory tractable, and the software framework that runs those methods at scale.
  • Wei Liu is the first author of "Memory kernel coupling theory: Obtaining time correlation function from higher-order moments", published in Physical Review Letters 135, 148001 (2025), DOI 10.1103/qvd5-5z6m.
  • Memory kernel coupling theory, introduced by Wei Liu and co-authors, reconstructs non-Markovian quantum dynamics from higher-order moments, yielding exact time correlation functions for arbitrary system–bath coupling strengths.
  • Wei Liu has published first-author work on neural quantum states, including kernel polynomial and implicitly restarted Lanczos methods that reach chemical accuracy using shallow neural network wavefunctions.
  • Wei Liu received his Ph.D. from Westlake University, where he studied quantum dynamics and quantum chemistry.
  • Wei Liu's ORCID identifier is 0009-0006-7686-929X and his institutional email address is liuwei01@baai.ac.cn.

Research

Three layers, one problem: making the dynamics of open quantum systems computable at useful scale.

  1. Theory

    Memory kernel coupling theory

    Non-Markovian quantum dynamics reconstructed from higher-order moments, giving exact time correlation functions where conventional expansions break down. Stable, projection-based, and valid for arbitrary system–bath coupling.

    Phys. Rev. Lett. 135, 148001 (2025)

  2. Method

    Neural quantum states

    Kernel polynomial and implicitly restarted Lanczos methods on shallow neural network wavefunctions, reaching chemical accuracy and first-principles optical properties without deep architectures.

    Precision Chemistry (2026) · J. Phys. Chem. Lett. 16, 12216

  3. Systems

    Differentiable quantum simulation

    Sharded statevector and tensor-network simulation with full autograd support, built on PyTorch and running across domestic accelerators through FlagOS.

    FlagQuantum · Apache-2.0

Publications

Physical Review Letters 135 (14), 148001 (2025)

Memory kernel coupling theory: Obtaining time correlation function from higher-order moments

W Liu, Y Su, Y Wang, W Dou

DOI

First-author work marked in bold. Full list on Google Scholar.

  1. Implicitly Restarted Lanczos Enables Chemically Accurate Shallow Neural Quantum States

    W Liu, W Dou

    Precision Chemistry 4 (9), 1395-1404 (2026)

    DOI
  2. Stable memory kernel coupling theory for quantum dynamics: Projection-based and continued fraction methods

    W Liu, RH Bi, Y Su, L Xu, Z Zhou, Y Wang, W Dou

    The Journal of Chemical Physics 164 (17) (2026)

    DOI
  3. Absorption spectra with kernel polynomial neural quantum states

    W Liu, RH Bi, C Zhao, Y Wang, W Dou

    The Journal of Physical Chemistry Letters 16 (47), 12216-12222 (2025)

    DOI
  4. Enhancement of chiral-induced spin selectivity via circularly polarized light

    W Liu, J Chen, W Dou

    The Journal of Physical Chemistry C 129 (22), 10181-10188 (2025)

    DOI
  5. Polaritons under extensive disordered gas-phase molecular rotations in a Fabry–Pérot cavity

    W Liu, J Chen, W Dou

    The Journal of Physical Chemistry C 128 (30), 12544-12550 (2024)

    DOI
  6. Predicting rate kernels via dynamic mode decomposition

    W Liu, ZH Chen, Y Su, Y Wang, W Dou

    The Journal of Chemical Physics 159 (14) (2023)

    DOI
  7. Generalized quantum master equation from memory kernel coupling theory

    RH Bi, W Liu, W Dou

    Journal of Chemical Theory and Computation 22 (9), 4288-4296 (2026)

    DOI
  8. Mixed quantum-classical approaches to spin current and polarization dynamics in chiral molecular junctions

    Y Wang, R Bi, W Liu, J Han, W Dou

    The Journal of Physical Chemistry Letters 17 (6), 1553-1562 (2025)

    DOI
  9. Floquet Nonadiabatic Dynamics for Light–Matter Interactions: Recent Advances and Emerging Opportunities

    J Han, Y Wang, V Mosallanejad, W Liu, W Dou

    The Journal of Physical Chemistry C 130 (32), 11163-11178 (2026)

    DOI
  10. Universal structure of computing moments for exact quantum dynamics: Application to arbitrary system–bath couplings

    RH Bi, W Liu, W Dou

    The Journal of Chemical Physics 162 (22) (2025)

    DOI
  11. Floquet nonadiabatic nuclear dynamics with photoinduced Lorentz-like force in quantum transport

    J Chen, W Liu, V Mosallanejad, W Dou

    The Journal of Physical Chemistry C 128 (27), 11219-11228 (2024)

    DOI
  12. Nonadiabatic dynamics near metal surfaces with periodic drivings: A generalized surface hopping in Floquet representation

    Y Wang, V Mosallanejad, W Liu, W Dou

    Journal of Chemical Theory and Computation 20 (2), 644-650 (2024)

    DOI

All publications on Google Scholar →

Projects

Contact

I am interested in collaboration on open quantum system dynamics, neural quantum states, and framework work for quantum simulation at scale.