Slide 1

Slide 1 text

© LY Corporation Terry So Machine Learning Engineer, DSG5ML1-Sol4 Team, DS Group, LY Corporation 2024.11.19 Introduction to Works of ML Engineer in LY Corporation

Slide 2

Slide 2 text

© LY Corporation Agenda 01 Self Introduction 02 Motivations to Work in Japan and LY Corp 03 Overview of Our Team 04 Current Works 05 Advices for Job Hunting on Data Expert Positions 06 Fresh Graduate Recruit 07 Contact Me 08 Q & A 2

Slide 3

Slide 3 text

© LY Corporation Self Introduction 3

Slide 4

Slide 4 text

© LY Corporation My Profile 4 2024 – present: Machine Learning Engineer, DSG5ML1-Sol4 Team, DS Group, LY Corporation 2019 - 2023: Machine Learning Engineer, LFK Data Labs, LINE Fukuoka 2018 - 2019: Data Scientist, LFK Data Labs, LINE Fukuoka 2013 - 2018: CRM Data Analyst, Data Service Department, Animals Asia Foundation Limited 2015 - 2017: MPhil in Data Science, Kansai University 2010 – 2013: BSc in Mathematics (Statistics), The Hong Kong University of Science and Technology Nationality: Hong Kong Terry So

Slide 5

Slide 5 text

© LY Corporation Hobbies and Interests 1. Drama 2. Movies 3. Pokémon video games 4. Stock trading 5. GYM 5

Slide 6

Slide 6 text

© LY Corporation Motivations to Work in Japan and LY Corp 6

Slide 7

Slide 7 text

© LY Corporation 7 Reasons to Work in Japan • Like Japanese Cultures ➢J-pop ➢Drama ➢Movie ➢Manga ➢Anime ➢Game etc.

Slide 8

Slide 8 text

© LY Corporation 8 Reasons to Join LY Corp 1. Large-scale data 2. Plenty of computational resources and infrastructures • Clusters of Servers • GPUs etc. 3. Able to learn and implement the most advance technologies 4. Able to engage in challenging and high-difficulty work. 5. Diverse working environment

Slide 9

Slide 9 text

© LY Corporation Overview of Our Team 9

Slide 10

Slide 10 text

© LY Corporation Team Structure 10 DS Group DSG5 ML1 Sol4

Slide 11

Slide 11 text

© LY Corporation 11 Roles of DSG5ML1-Sol4 Team 1. Develop ML/AI systems or services which can . ➢eg: 1. Natural Language Processing (NLP) 2. Computer Vision (CV) 3. Time Series Forecasting 4. Recommendation System etc. 2. Maintain those systems or services after deployed to production environment

Slide 12

Slide 12 text

© LY Corporation General Data Lake Pipeline Image source: https://cloud.google.com/blog/topics/developers-practitioners/what-data-pipeline-architecture-should-i-use/

Slide 13

Slide 13 text

© LY Corporation Current Works

Slide 14

Slide 14 text

© LY Corporation 14 Projects 1. General Recognition System 2. Yahoo Ad Metadata Generator

Slide 15

Slide 15 text

© LY Corporation 15 Line Stickers Review 1. Copyright violation • Duplicate stickers • Stickers containing copyright-protected ACGN (anime/comic/game/novel) characters 2. People Image right violation • Celebrities in particular countries eg. ➢ Royal family in Thailand ➢ Sensitive Politicians in Korea etc. 3. Some symbols prohibited by law in particular countries • Rainbow, Nashi, Rising sun flag

Slide 16

Slide 16 text

© LY Corporation Line Stickers Review Process Sticker Contents submitted by Users Sticker Contents Review Sticker Contents Approve / Reject

Slide 17

Slide 17 text

© LY Corporation Example: Anime Character Recognition Gallery: Query: score 0.9 0.5 0.3

Slide 18

Slide 18 text

© LY Corporation Example: Celebrity Recognition Gallery: Query: score 0.9 0.3 0.1 score 0.89 0.4 0.2

Slide 19

Slide 19 text

© LY Corporation General Recognition System Detection Embedding Searching Anime Character Recognition Celebrity Recognition

Slide 20

Slide 20 text

© LY Corporation 20 Projects 1. General Recognition System 2. Yahoo Ad Metadata Generator

Slide 21

Slide 21 text

© LY Corporation 21 Raw Yahoo Ad-submission Data • Original text data for ad oExample (dummy data for demo): ➢{‘advertiser account name’: ‘terry abc corp’, ‘campaign name’: ‘terry abc ad campaign’, ‘advertisement title’: ‘terry abc ad 101’, ‘advertisement description’: ‘1st advertisement of terry abc corp’, ‘advertisement link url’: ‘www.terrysocorp.com’} ❖ Missing: • Goods? • Categories of Good? • Brands of Good? • Company Name? • …

Slide 22

Slide 22 text

© LY Corporation 22 Raw Yahoo Ad-creative Data • Original Image data for ad oExamples (dummy data for demo): ➢1. ➢2. If you want Cosplay make up service, find Terry abc Cosplay costume

Slide 23

Slide 23 text

© LY Corporation 23 Yahoo Ad Metadata Generator Example(dummy data for demo): ➢ {‘Goods’: ‘cosplay make-up, costume services’, ‘Categories of Goods’: ‘clothes and make-up’, ‘Brands of Goods’: ‘terry cosplay’, ‘Company Name’: ‘terry abc corp’, …}

Slide 24

Slide 24 text

© LY Corporation Why Generate Ad Metadata?

Slide 25

Slide 25 text

© LY Corporation Advices for Job Hunting

Slide 26

Slide 26 text

© LY Corporation What You Can Do Now 1. Be proficient in at least one programming language ❖ Python IDE: ➢Jupyter Notebook ➢Visual Studio Code

Slide 27

Slide 27 text

© LY Corporation What You Can Do Now 2. Be proficient in relational database ❖ MySQL ➢Tutorial

Slide 28

Slide 28 text

© LY Corporation What You Can Do Now 3. Be proficient in Github • Utilize opensource resources (eg. data, models, codes etc.) contributed by others • Share your own creation to public

Slide 29

Slide 29 text

© LY Corporation What You Can Do Now 4. Join Kaggle • Kaggle Competition provides many useful industry- level resources: ➢Data, ➢State-of-art Models, ➢latest data science techniques and technologies ➢Common approaches in industry fields

Slide 30

Slide 30 text

© LY Corporation What You Can Do Now 5. Register LinkedIn account and prepare your profile • Get latest job market information • Promote yourself and your expertise • Connect people in ML/data science community

Slide 31

Slide 31 text

© LY Corporation What You Can Do Now 6. Register OpenAI account and be proficient in Chatgpt • Answer your questions about AI/ML and job hunting immediately • Rapidly start your coding • Help generate some data you need • Get sense of what generative AI and Large Language Model (LLM) are • Know trend of AI/ML fields

Slide 32

Slide 32 text

© LY Corporation Join Us Fresh Graduate Recruit • Engineer (data science course): ➢https://www.lycorp.co.jp/en/recruit/newgrads/engineer/

Slide 33

Slide 33 text

© LY Corporation Contact Me LinkedIn: www.linkedin.com/in/wai-tik-so-6563b3138