TAICHANG ZHOU

***, ******** ********* *****, ******** California, United States

424-***-**** ad1td4@r.postjobfree.com

EDUCATION

University of California, Los Angeles Sep 2023 - May 2025 (expected) M.S. in Computer Science

Hong Kong University of Science and Technology Sep 2019 - May 2023 B.Eng. in Computer Science and Engineering Overall GPA: 3.7/4.3(top 10%) Georgia Institute of Technology Jan 2022 - May 2022 Exchange Overall GPA: 4.0/4.0

SKILL-SET

Programming Language: C/C++, Python, Java, HTML, CSS, JavaScript(JS), SQL Machine Learning: Numpy, Pandas, Pytorch, CNN, Transformer, LSTM, Large Language Model(LLM) Other: Cloud Computing(HDFS,Hadoop,Spark), Recommendation System, Blockchain INTERNSHIP

Vetrackr Jun 2022 - Aug 2022

Junior AI Developer Hong Kong

· Collaborated with the team to contribute to risk management model of predicting taxi drivers’ security ranks

· Retrieved the weather data through Python web crawler, and collected driving data through plug-in detectors on taxis

· Implemented algorithms based on GPS to identify distinctive driving behaviors with over 90% accuracy.

· Established a deep learning system from scratch with multi-layer LSTM to predict driver risk appetite and recommended speed based on weather conditions, attaining a precision over 40%

· Evaluated the risk prediction model with saliency analysis to ensure the reliability and robustness PROJECT

Glancee - Adaptive Learning System Sep 2020 - Jul 2021 Co-Author Accepted by CHI ’22: Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems

· Constructed an adaptive learning system for instructors to grasp student learning status in synchronous online classes

· Performed crapping data, pre-processed the data including data cleaning and data aggregation, and labeled the dataset of emotion-expression pair videos

· Assisted in implementation of Openface to identify 68 face landmarks, and emotion/engagement detectors to identify stu- dents’ emotions from facial landmarks

· Implemented front-end page for 3 views of the system by HTML, JS, and CSS: dashboard view (outlook adjustment), in-class view (status collection from student and illustration for lecturer) and post-class view (review for lecturer) Moral-LLM Otc 2023 - Present

· Reviewed Parameter-Efficient Fine-Tuning(PEFT) methods that can reduce the GPU memory usage, including QLora, LoRa, Adapter, etc.

· Combining QLoRa with another PEFT technology, Aligner. ProtoMed - Trustworthy Medical Assistant Sep 2022 - May 2023

· Led the team to research on Explainable AI(XAI) methods to identify insights about the inner unit in deep learningmodels used for breast MRI and skin diseases to generate trustworthy and interpretable model

· Built the training set by web-scraping the medical data from different hospital websites for further model training, prepro- cessed the data with SITK/Pandas/Numpy library

· Formulated the DenseNet121 and ResNet50 models, pruned to AUC above 70% for 114 categories and matched Convolu- tional Layers with medical concepts by TCAV

· Pruned themodelwithoptimizerAdamWandconstructedConceptCorrectionlayertoensuresignificantclinicalconcepts retrieved by TCAV were learned

· Introduced explainability to deep learning model but retained the performance Thursday - Recipe Recommendation Feb 2022 - May 2022

· Designed and developed a transparent Hybrid user-adjustable recipe recommendation systemwithcontent-based and user- collaborative filtering. The user can customize those thresholds by themselves

· Constructed the content-based model by converting the raw text data to vectors through the Word2vec algorithm and weighting the ingredients in the recipe with TF-IDF scores

· Computed the cosine similarity for input and filtered the top 50 results based on the user customized filters

Contact this candidate