• Utilize Reinforcement Learning from Human Feedback (RLHF) to improve reasoning of the current state-of-the-art machine learning models.

• 用Python從零開始開發GPU和CPU版本的混合蒙特卡羅程式，利用基於雙譜描述符的機器學習原子間勢進行能量評估，與CPU版本相比，GPU運行代碼使計算速度提高了10倍。模擬與現有的實驗結果相符，並解釋了先前研究中所缺少的奇數和偶數2D鈣鈦礦成員之間的差異； # results from the simulations were presented at the Materials Research Society (MRS) Conference in Singapore (05-2023) and International Conference on Molecular Simulation (ICMS) in Taipei, Taiwan (10-2023); • 採用ML-AIMD （VASP）和基於大規模機器學習的分子動力學（Matlantis）來證實用於電池陽極的氟化修飾納米石墨烯的鋰存儲能力增強，與實驗結果一致。此外，這些模擬還對鋰的優先吸附位點提供了全新的原子級別的見解，這是當前的實驗技術無法實現的； • 對摻雜的無機雙鈣鈦礦進行基於VASP的超級電池優化和能帶結構計算，所得結果與實驗資料完全一致，並解釋了所觀測到的間接帶隙； * all calculations carried out on Taiwania I, II and III supercomputers; * National Science and Technology Council (NSTC) postdoctoral fellowship --- @ Computational Nanomaterials & Nanomechanics Lab

• Developed a Grand Canonical Monte Carlo (GCMC) 2D perovskite crystal growth simulation code in Python programming language, employing the machine-learned potential, leveraging PyLammps library as an interface to LAMMPS molecular simulation package for energy evaluations of the supercell structures during the trial moves; • Trained the machine-learning enabled potential for butylammonium (BA) based 2D perovskite material employing the Behler-Parrinello feed-forward neural network (FFNN) method based on the Gaussian descriptor as implemented in Atomistic Machine-learning Package (AMP); • Assembled a training set from ab-initio molecular dynamics trajectories (AIMD) including the most important inter-atomic contributions in 2D perovskite material as an input for the subsequent machine learning of the artificial neural network (ANN) force-field; • Carried out the surface structure of low-index facets in 2D Ruddlesden-Popper material from density functional theory (DFT) calculations, - extracted surface features revealed that low surface energies stem from hydrogen bond shortening increasing the strength of Coulomb interaction between organic cations and iodine atom; # results presented at the WCCM ECCOMAS Conference in Paris, France (01-2021); • DFT study of adsorption of small organic molecule on lead-free antimony-based all-inorganic perovskite, - calculations revealed noticeably stronger affinity of O atom within the N-methylpyrrolidone (NMP) molecule to Cs atom in all-inorganic Cs3Sb2I9 perovskite compared to S atom within thiourea (TU) leading to better morphology and increased stability of the active layer of the solar cell device; • Carried out large scale DFT study of intertwined defect in 2D layered Ruddlesden-Popper 2D perovskite material in VASP, - extracted the band structure and density of states (DOS) of the resulting trajectories reveal shallow defects and Fermi level pinning features in the material; * all calculations carried out on Taiwania I, II and III supercomputers; --- @ Computational Nanomaterials & Nanomechanics Lab

• Carried out metadynamics simulations of methylammonium lead-iodide (MAPbI3) soft crystal using DL_POLY Classic molecular dynamics simulation software; - extracted the free energy surface (FES) and the energy barrier for crystal nucleation from amorphous phase at different temperatures; # results including the energy barriers and critical nuclei presented at the Association of Computational Mechanics Conference (ACMT) in Tainan, Taiwan (10-2017)}; • Carried out the Temperature Accelerated Dynamics (TAD) simulations of cation defect in MAPbI3 crystal in LAMMPS Molecular Dynamics Simulator, - results revealed different diffusion barrier and cation pathways with and without applied external electric field; • Molecular dynamics simulations of methylammonium lead-iodide (MAPbI3) soft crystal, performed in LAMMPS molecular dynamics simulation software - to test the phase stability and temperature limits of the MAPbI3 interatomic potential; * all simulations carried out on local cluster including 16 core and 12 core notes --- @ Computational Nanomaterials & Nanomechanics Lab

• Commenced experiments to investigate cellular responses in hypoxic environment generated by spatially controlled chemical reactions within the PDMS-based microfluidic device; • Performed the simulation tests of the flow regime in the microfluidic device using COMSOL Multiphysics simulation package; • Fabricated PDMS-based microfluidic devices for hypoxia study; • Cell culture - passaging cells on daily basis; @ Lab for Integrated Biomedical Microdevices

• Lab rotation @ Molecular and Material Modeling Lab [Institute of Atomic and Molecular Sciences] (02-2015 - 04-2015), - literature review of available classical H2O force-field models (TIP3P, TIP4P) to study phase transition in water from molecular dynamics simulations; • Lab rotation @ Polymer Physics and Complex Fluids Group [Institute of Physics] (09-2014 - 01-2015), - literature review of DNA in nanoconfinement from molecular dynamics simulations study, - literature review and project proposal for polymer-based solar cell active layer material study from molecular dynamics simulations;

• 用Microsoft Visual Studio開發了一個獨立的聚苯乙烯間歇式反應器應用，該應用集成了C#前端和Fortran後端，從而可以改善基於用戶輸入的反應器輸出的模擬； • 用C#設計了一個符合CAPE-OPEN標準的聚苯乙烯間歇式反應器單元操作模組，使IT數學模型在行業標準流程類比軟體（如：Aspen Plus, COCO Simulator）中的應用成為可能； • 開發了一個用於聚苯乙烯間歇式反應器參數輸入的VBA/Excel介面，作為中間工作模型，成功地提交給巴斯夫公司； * Under the 'Multimod' [Multiscale modeling] project hosted by the Aristotle University of Thessaloniki, Marie Curie fellowship

• 用SMath Studio簡化了原料配比、隧道式乾燥器和爐體質量/熱量傳遞的計算，使輸出計算時間減少10倍； • 對隧道式乾燥器進行基於鐳射測量儀的物理測量，為後續的軟體視覺化和數學建模提供必要的資料； • 每天控制原材料和最終產品的品質；

• 操作奧林巴斯血液分析儀對血清進行分析，管理樣本，並對結果進行分類；

•負責臺式電腦的維護：作業系統/應用的安裝、硬體/局域網的故障排除以及維修； • 使用桌面出版軟體解決方案為學院首席研究員（PI）提供PPT演示並準備書籍章節；