AI4S PhD researcher (Peking University) in computational catalysis & machine-learning potentials. I ship research-grade atomistic workflows that feel like production software — opinionated defaults, reproducibility, good UX for scientists and AI agents — around the DeepModeling / ABACUS ecosystem.
- Homepage: https://quantummisaka.github.io/
- ORCID: https://orcid.org/0009-0007-8286-9090
- Email:
quanmisaka [at] stu.pku.edu.cn
- Research: iron-based Fischer–Tropsch & surface catalysis; ML potentials via large-atomic-model fine-tuning and active learning.
- Engineering: reproducible, agent-friendly pipelines — DP‑EVA (active learning), ATST‑Tools (transition-state workflows), ABACUS toolchain.
- Community: ABACUS / DeePMD open-source maintainer & community operator; project leader at Beijing Sidereus Intelligent Computing Technology Co., Ltd. (Sidereus AI).
| Project | What it is |
|---|---|
| DP‑EVA | Data-efficient active-learning fine-tuning of DPA universal potentials. |
| ATST‑Tools | Governed, YAML-driven ASE transition-state toolkit for ABACUS & DeePMD. |
| ABACUS user guide | Chinese documentation & onboarding for ABACUS. |
| ABACUS toolchain | Cross-platform ABACUS installation automation (major contributor). |
DP‑EVA — Clean Energy (2026, first author) · FT²DP — J. Mater. Inform. (2025, first author) · 7 papers on ORCID
Building agentic tools for materials simulation — planning, verification, automation.