Marsan Ma
Staff Data Scientist
Tokyo, Japan
Summary
👤
Senior🎓
Top SchoolMarsan Ma is a Staff Data Scientist based in Tokyo with nearly a decade of experience applying machine learning and full-stack engineering to high-impact product problems. He has driven billion-dollar business improvements at Indeed by building user behavior modeling, recommender, and bidding systems that meaningfully improved conversion, cost-per-action, and recommendation diversity. Marsan blends deep learning (BERT, Transformers, TextCNN), graph and siamese-CNN approaches for cold-starts, and pragmatic ML engineering—owning data pipelines, daily retraining, and cloud infra across AWS/GCP/Azure. He also has hands-on experience in productionizing seq2seq chatbots (implementing beam search with an anti-language model), fraud detection, and real-time dashboards that informed business decisions. A mentor and technical interviewer, he pairs strong research instincts from an NTU EDA background with a track record of shipping robust, revenue-driving systems.
9 years of coding experience
8 years of employment as a software developer
Master's degree, Electronic Design Automation, A, Master's degree, Electronic Design Automation, A at National Taiwan University
Chinese, english (toeic 910), japanese (n2 certificate), Mandarin
Github Skills (4)
tensorflow10
language-model10
python10
deep-reinforcement-learning5
Programming languages (2)
Github contributions (5)
Marsan-Ma-zz/tf_chatbot_seq2seq_antilm
Sep 2016 - Mar 2017
Seq2seq chatbot with attention and anti-language model to suppress generic response, option for further improve by deep reinforcement learning.
Role in this project:
ML Engineer
ML EngineerContributions:27 commits in 6 months
Contributions summary:Marsan implemented beam search and an anti-language model (anti-LM) into the seq2seq chatbot. These additions aimed to suppress generic responses and improve the quality of generated text. The code modifications focused on adjusting probabilities within the beam search algorithm, likely to enhance the chatbot's ability to generate more relevant and coherent responses.
generic-responselanguage-modelantiseq2seqseq2seq-model
Marsan-Ma-zz/leetcode-python-sols
Sep 2016 - Jan 2017
Contributions:36 commits in 3 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.