Lin Geyu | ML Engineer & Researcher

About Me

Python 90%

Pytorch 90%

Machine Learning 90%

Large Scale Model Training 80%

Java 70%

C++ 70%

Tensorflow 70%

SQL 70%

Spark 60%

ReactJS 50%

SpringBoot 50%

I am a Computer Science student in National University of Singapore graduated in 2023. The skills section shows the skills that I know and my confidence level.

I am enthusiastic about developing machine learning application in different domain. I have done several machine learning projects with respect to Human Activity Recognition, Image Classification and Planning using Reinforcement Learning. I also have background about developing full stack application and system testing for large scale system.

I am interested in discussion about Machine Learning techniques and cutting edge application development technique.

Work Experience

Professional Experience

Data Scientist

OCBC

2025 - Present

Central AI service

Research Engineer

Institute for Infocomm Research, Agency of Science Technology & Research

2023 May - 2025 Oct

Customize distributed training pipeline which utilize mosaic, pytorch and transformers library to customize distributed behavior which scales the training to 320 H100 GPU cards and train the MERaLION 2 model, also finetuning the whisper towards Singapore local context.
Structured 200k hours open-sourced speech data into uniformed huggingface datasets format to support training of large speech foundation model.
Research the methodologies of enhancing large language model to improve the multilingual consistency of the models using instruction finetuning.
Build Pytorch model training pipeline to improve speech evaluation for unseen data by introducing objective metric combining with human label.

Data Science and Machine Learning Intern

Agoda

2022 Aug - 2023 May

Design the system to support online reinforcement learning
Implement general Reinforcement Learning framework with different exploration algorithm
Adapt different use cases with respect to the framework

Research Intern

NUS Ubicomp Lab

2020 Dec - 2022 May

Develop Android Application to support data collection from Esense earbuds device (Accelerometer, gyroscope)
Collect face touching activity data using earbuds sensor for 30 participants
Implement 1DCNN, LSTM model for activity recognition (classification)

Software Engineer Intern

PayPal

2021 Jan - 2021 July

Develop Internal Issue Tracker using ReactJS, Springboot and Mongo DB
Design API for website to fetch issue information from JIRA
Support Auto email of the summary and analysis of issue
Improve regression testing by creating java aspect to precreate data and fetch data directly during test time
Reduce regression test time by 40%