Jack Chih-Hsu Lin, Phd Email & Phone Number

California, United States

• Towards Data Science is one of the largest data science publications (650K followers).
• Mastering GenAI ML System Design Interview: Principles & Solution Outline (2024): https://towardsdatascience.com/mastering-genai-ml-system-design-interview-principles-solution-outline-71a4664511a7
• Mastering GenAI ML System Design Interview (2): Design ChatGPT Memory Feature (2024): https://medium.com/towards-data-science/mastering-genai-ml-system-design-interview-2-design-chatgpt-memory-feature-fe908517d76c
• Scaling Monosemanticity: Anthropic’s One Step Towards Interpretable & Manipulable LLMs (2024):https://medium.com/towards-data-science/scaling-monosemanticity-anthropics-one-step-towards-interpretable-manipulable-llms-4b.
• Optimize PyTorch Performance for Speed and Memory Efficiency (2022) (10k+ views in a week): https://towardsdatascience.com/optimize-pytorch-performance-for-speed-and-memory-efficiency-2022-84f453916ea6
• Self-Supervised Learning (SSL) Overview : https://towardsdatascience.com/self-supervised-learning-ssl-overview-8a7f24740e40

San Francisco Bay Area

• Develop/deploy agentic RAG framework (reflection, memory) for querying structured data via natural languages
• Pre-train and fine-tune Code Llama, Llama 3, StarCoder2 with LoRA on multi-GPUs for text-to-database queries
• Develop/deploy agentic automatic synthetic data generation pipeline, boosting model performance by 15%
• Align the LLMs with human preference by chain of hindsight to reduce hallucination

San Francisco Bay Area

• Write weekly reports to executives, work with product manager/designers, engineers and explain DS approach to customers/stakeholders for the project
• Developed multi-task neural networks for hierarchical forecasting of business deals and company revenue
• Developed large-scale deep learning model trained with a graph of 3M nodes and 33M edges
• Built/Deployed the algorithm pipeline (reducing cost by 500000x) on 100 workers to determine spatial temporal correlation among 9.8B time series geospatial data points and generated a heterogenous knowledge graph.
• Improved the recall of diagram parsing (computer vision, object detection) by 20-80% using Faster R-CNN in Keras
• Analyzed time series geospatial data of 800M records by distributed computation (batch jobs and MapReduce)

Houston, Texas, United States

I invent and implement a new type of neural network that reduces 200 times of parameters and converges 35% faster in PyTorch. I achieve AUROC 0.88 while predicting therapeutic target genes by modeling high dimensional (~10,000) data in human cancer cell lines. I analyze and validate the prediction by statistical tests.

Houston, Texas Area

I merged and cleaned data from 3 databases and generated a network of 215,000+ drug-gene-disease associations. I implemented and validated graph-based diffusion in Python to predict associations which were proven to be true either in later database releases or literatures with >90% precision. The results were published as a 1st-author paper in.

Texas

I led a team of 30+ persons from 9 institutes/colleges in 4 Texas cities to facilitate intellectual conversation and networking among Taiwanese professionals in bioscience field worldwide. I organized the annual symposium in 2019, and the participant number increased by 27%.

2019/10/02 Gave a talk titled "One of the Best Ways to Explore Your Career: the Internship, a Thrilling Journey to Illumina iAspire" at Texas Taiwanese Biotechnology Association Webinar, Houston, Texas. 2019/08/04 Gave a talk titled "Texas: the Third Coast or the Third World of Biotech?" at Boston Taiwanese Biotechnology Association 2019 Annual Symposium.

I am overall ranked top 3.0% (5,484/179,945). I have won top 3.0% (26/866), top 4.9% (188/3,835), and top 6.1% (201/3,274) in 3 competitions (in image and tabular data) using data science, machine learning and deep learning skills: data cleaning, missing data imputation, exploratory data analysis, statistical analysis, feature selection/engineering.

Greater San Diego Area

I analyzed clinical sequencing data to diagnose patients with rare and undiagnosed genetic diseases. I decreased 94% of time spent in manual annotation of colleagues by developing a new machine learning pipeline. I found unexpected, creative and meaningful features to improve predictions. I collaborated with multiple teams to customize the pipeline to.

Greater San Diego Area

I organized meetings and led a 11-intern cross-functional team to win the 1st place out of 10 teams in the business case competition. We developed strategies and solutions for a challenge Illumina was facing by using everyone's expertise. The team consisted of people from diverse departments e.g., marketing, financing, engineering, computational biology.

I processed hundreds of gigabytes of 5 types of cancer genomics data using Perl and R.

I analyzed RNA-seq data of 70 million pair-ended reads from 12 yeast samples using Perl and R

Doctor Of Philosophy (Phd), Quantitative Computational Biology, Gpa 4.0

Activities and Societies: Texas Taiwanese Biotechnology Association, Taiwanese Student Association, Volleyball, Badminton.

Master Of Science (Ms), Plant Pathology And Micriobiology, Gpa 4.0

Bachelor'S Degree, Plant Pathology And Microbiology