Chih-Hao Liu

I am Steve, a dedicated professional in AI Research, 5G technology, and Materials Sciences. I enjoy street dancing, channeling my creativity into writing blogs, listening to podcasts about international affairs, and exploring new destinations through travel.

  Taipei, Taiwan         

Working Experience


Research Assistant (part-time) – 5G technology 

Institute for Information Industry 

05 2022 - 08 2023
Taipei, Taiwan

  • Analysis specification: 3GPP 5G NR/4G LTE L1/L2/L3, O-RAN 11 WG, ETSI MANO. 
  • Source code mapping: O-RAN open source project (RIC/SMO/O-Cloud/xAPP), ONAP (SDNR/CCSDK), and OAI (L1/L2/L3). 
  • Market research: telecom operators open RAN product & solutions, O-RAN integration & testing, and white paper reading. 

Research Assistant (part-time) – AI in Medicine 

National Taiwan University

08 2021 - 08 2023
Taipei, Taiwan

  • Collaborated with a biotech startup, hospital, and research institutes to develop AI tool and applications on human skin. 
  • Data management, data storage, image processing, DL/CV algorithms design, SOTA model comparison, model fine-tuning, model optimization, model management, quality assessment, IRB report, journal, and patent writing. 
  • Authored 10 publications (1 st author × 8 / 2nd author × 2) with 1 published, 3 under review, 6 manuscripts completed and 1 patent (applied for provision). 

Programming Lecturer 

(part-time)

Coding APE Ltd.

09 2021 - 08 2023
Taipei, Taiwan

  • Teach teenagers about data structures and algorithms, advanced placement, computer science, C++/Python object-oriented programming, web crawler, Line chatbot design, Pygame development.

Research Assistant (full-time) – Physics 

National Taiwan University

09 2020 - 02 2021
Taipei, Taiwan

  • Researched in non-destructive optical inspection for measuring the thermal properties of semiconductor materials. 
  • Designed optical experiments, operated equipment, and system signal processing, and conducted materials analysis. 
  • Published in Taiwan Thermoelectric Conference and honored with the NTU CCMS Innovative Competition Excellent Award. 

Summer Research (intern) - Mathematical Theory 

National Center for Theoretical Sciences

07 2020 - 0 2020
Taipei, Taiwan

  • Researched in time series analysis using integral transform and machine learning. 
  • Studied the mathematical theory on scattering transform and dimensional reduction for signal processing. 

Education


National Taiwan University, 

M.S. in Electrical Engineering and Computer Sciences -  Optoelectronics  •  2021 - 2023

  • [Thesis]: Quantitative and Qualitative Analysis of Human Skin Structures and Lesions in Cellular-Resolution OCT Images by Deep Learning (English version, 408 pages).
  • [CS Courses]: Algorithms Design and Analysis (audit), Operating System, Compiler Design, Applied Deep Learning, Data Structures (audit), Database Management. 
  • [EE Courses]: Medical Photonics, Silicon Photonics, Integrated Optics, IC Engineering, Optoelectronics, Solid State Lighting, Semiconductor and Display, and Optical Techniques in Diagnosis, Optical Simulation, Optical System. 

National Cheng Kung University

Double Degree in B.S. in Optical Science and B.Eng. in Chemical Engineering  •  2021 - 2023

(Studied Civil Engineering during 2015 to 2017, Transferred to Optical Science at 2017)  

  • [Courses]: Take more than 240 credits and more than 100 courses, including Quantum Physics, Molecular Simulation, Process design, Big Data Analysis, Finite Element Analysis, Optical Communication.
  • [Activity]: Participate in 3 student clubs
  • [Research]: Conduct undergraduate research in 3 distinct labs and 3 side projects in senior/graduate level courses. 

Skills


Software Engineering  

Python, PyTorch, Tensorflow, SciKit learn, Hugging Face, OpenCV, C/C++, Java, MySQL, PostgreSQL, Spark, Hadoop, BS4, Selenium, GCP, BigQuery, Flask, HTML, CSS, JS, LabVIEW, Linux, Ubuntu, Kubernetets, Git/Github. 

Computer Sciences

Computer Programming, Operating System, Compiler Design, Applied Deep Learning, Algorithm Design & Analysis (audit), Data Structures (audit), Computer Architecture (audit), Database, Statistics, Big Data Analysis.

Machine Learning 

XGBoost, DL, CV, NLP/NLU, RL, XAI, MLOps, Recommendation Systems, LLM Fine-tuning & Prompting, Self-supervised Learning, Few-Shot Learning, SHAP, Grad-CAM, ETL, EDA, MapReduce, DB Management. 

Mathematical Simulation 

MatLab, ANSYS, COMSOL, R , Numerical Simulation, Numerical Analysis, Finite Element Analysis, Differential Equation (Eng. Mat. I), Linear Algebra (Eng. Mat. II), Fourier Analysis (Eng. Mat. III), Probability 

Communication Specification

3GPP 5G NR/4G LTE spec, L1/L2/L3 architecture, ETSI MANO framework, O-RAN spec, OSC, ONAP.

Soft skills

Problem-Solving, Strong Adaptability, High Flexibility, Patience, Good Verbal Communication, High Learning agility 

Honor and Award


  • Hahow Ltd. Course Recommendation System Design Competition – Mentioned Award / Kaggle Rank 1. (May, 2023). 
  • National Synchrotron Radiation Research Center Winter School – Group Presentation 1st Prize. (Jan, 2022).
  • NTU Center for Condense Matter Sciences Innovation Competition – Excellent Award. (Feb, 2021)
  • NCKU Graduate Representative (Jun, 2020) 

Publication and Patent 


Journal paper

  • [1] Tsai, S. T.,Liu, C. H., Chan, C. C., Li, Y. H., Huang, S. L., & Chen, H. H. (2022). H&E-like staining of OCT images of human skin via generative adversarial network. Applied Physics Letters, 121(13), 134102. 
  • [2] Liu, C. H., Fu, L. W., Chen, H. H., & Huang, S. L. (2023). Toward cell nuclei precision between OCT and H&E images translation using signal-to-noise ratio cycle-consistency. Computer Methods and Programs in Biomedicine, 107824. 
  • [3] Liu, C. H., Fu, L. W., Chang, S. W., Wang, Y. J., Wang, J. Y., Wu, Y. H., Chen, H. H., & Huang, S. L. (2023). Optimization and limitation of U-Net-based segmentation for quantitatively evaluating in-vivo human skin layers and keratinocytes in cellularresolution FF-OCT imaging. Journal of Biomedical Informatics (under review). 
  • [4] Fu, L. W., Liu, C. H., Jain, M., Chen, J. C. S., Wu, Y. H., Huang, S. L., & Chen, H. H. (2023) Training with uncertain annotations for semantic segmentation of basal cell carcinoma from FF-OCT images. IEEE Transactions on Medical Imaging (under review). 
  • [5] Liu, C. H., Chen, Y. S., Chen, Y. C., & Huang, S. L. (2023). Cellular-resolution OCT image denoising with unpaired reference guided deep generative network. Biomedical Signal Processing and Control (submitted). 
  • [6] Liu, C. H., Wang, Y. J., Wang, J. Y., Wu, Y. H., & Huang, S. L. (2023). 3D in-vivo human skin keratinocytes segmentation in FFOCT images from 2D pseudo label using self-training. Computerized Medical Imaging and Graphics (submitted). 
  • [7] Liu, C. H., & Huang, S. L. (2023). Unpaired in-vivo OCT and ex-vivo H&E human skin image translation using auxiliary learning. Medical Image Analysis (submitted) 
  • [8] Liu, C. H., Wang, Y. J., Wang, J. Y., Wu, Y. H., & Huang, S. L. (2023). Self-supervised in-vivo OCT human skin lesions classification with curriculum learning path. (manuscript). 
  • [9] Liu, C. H., Wang, Y. J., Wang, J. Y., Wu, Y. H., & Huang, S. L. (2023). Few-shot in-vivo human skin lesions identification in cellular-resolution FF-OCT image with contrastive pre-training. (manuscript). 
  • [10] Liu, C. H., & Huang, S. L. (2023). Monte Carlo OCT simulation assisting real world segmentation with few data. (manuscript).

Conference paper

  • [1] Liu, C. H., & Chang, C. W. (2021) Infrared frequency domain thermoreflectance measurements of thermal conductivity. Annual Meeting of Taiwan Themoelectric Society. 
  • [2] Liu, C. H., & Tseng, S. Y. (2019) Short and broadband mode-division multiplexers using shortcut to adiabaticity and supersymmetric optical structures. Optic and Photonic Taiwan International Conference.  

Patent

  • [1] Deep Learning-Based Conversion of Optical Coherence Tomography Image to Pathological Slice Image for Biomedical Structure and Composition Mapping. (US Provisional Patent).

Projects Experience 


O-RAN Open Source Project Study 

Conducted a source code mapping analysis for O-RAN 5G specifications, covering components such as Near-RT RIC/Non-RT RIC, SMO, and xAPP/rAPP. Analyzed E2/A1/O1/O2/R1 interface transport APIs and investigated the relationship between the 3GPP framework and O-RAN's 7.2x split architecture. Explored the ETSI MANO framework in the context of SMO/O-Cloud schemes.

2D/3D Human Skin Cell Segmentation Tool Development 

Collaborated with a biotechnology startup to develop 2D/3D segmentation models and enhance image quality. Established and optimized U-Net model training procedures, resulting in a significant accuracy improvement from 41% to 72%. Achieved acceleration in GPU inference speed from 3 minutes to 16 seconds, and reduced memory storage requirements by a factor of 8.

Mackay Memorial Hospital Lesions Identification Model Design 

Partnered with Mackay Memorial Hospital to design AI models for the clinical diagnosis of human skin conditions. Conducted EDA on a dataset comprising data from over 100 patients, achieving a 91% accuracy in identifying 14 different skin lesions. Developed a few-shot model capable of detecting diseases with limited data availability. Conducted research on visual explanation figures to gain insights into the model's decision-making process. 

Pseudo Pathological Slices Application Development 

Collaborated with NTU Hospital, NTU CSIE, and NTU GIEE to develop a generative AI model capable of translating non-invasive microscopic images into conventional H&E slices. Curated a diverse image dataset and designed novel DL algorithms that achieved SOTA performance. Enhanced translation accuracy through auxiliary learning and patent the concept. 

Hahow Ltd. Course Recommendation System Design 

Designed and implemented a ML-based recommendation system to predict courses likely to be purchased by both new and existing customers. Conducted a comprehensive EDA on customer data, leveraging multiple ML algorithms. Explored customer behavior and assessed the impact of advertising efforts. Skillfully managed feature engineering to optimize the performance. 

Personal Achievements 


Linux Programming Study Note Series 

Writing 25+ articles on Medium, providing the holistic introduction of system programming concept. Conduct hand-on developments for Linux system call and instruction using C programming language.

Machine Learning Study Note Series 

Writing 100+ articles on Medium, providing in-depth explanations of ML algorithms' theories. Implemented these algorithms from scratch, without reliance on ML toolkits, to ensure a strong foundational understanding. 

NCKU Freshman Lecture – Speaker 

Lecturing to over 300 incoming freshman students. Shared personal insights and experiences regarding major selection, double majors, minors, and the university's available resources, helping students make informed decisions about their academic paths. 

Optic and Photonic Taiwan International Conference – Student Event Organizer 

Coordinated a chapter event attended by 300+ participants, including international scholars and esteemed guests. 

Extracurricular Activities


President of NCKU Student Chapter / Optical Society of America (Jul, 2019 ~ Jun, 2020) 

Promote optical science and be the bridge between students and corporate. Organize the company visitation, research institute visitation, and professional lectures 

Group Member / Yunnan Overseas Services Learning Program (Sep, 2018 ~ Oct, 2019) 

build 10 eco-friendly toilets in Yunnan China Country Side. 

Choreographer / NCKU Pop Dance Club (Sep, 2017 ~ Jan, 2018) 

Dance Choreographing for the winter session presentation. 

Certification


Spark - Level 2

IBM

7fb1c8be-0569-4f66-871d-552ca9f6afcf

Hadoop Foundations - Level 2

IBM

22dc74ac-8924-4f55-ba3a-afd730a5c403

Perform Foundational Infrastructure Tasks in Google Cloud

Google

1115863

Perform Foundational Data, ML, and AI Tasks in Google Cloud

Google

1113540

Insights from Data with BigQuery

Google

1114498

Explore Machine Learning Models with Explainable AI

Google

1115556

Create ML Models with BigQuery ML

Google

1115438