đ
I am a Manager of AI at SKIM. I apply cutting-edge AI technologies, such as chatbots, image understanding and generation, to Market Research. In the past, I’ve worked on training LLMs at Huawei, and worked on NLP projects in two other internship projects.
Skills
Python
90%
PyTorch
90%
Git
90%
Bash
70%
Photography
60%
Rust
50%
Unity
10%
Experience
Manager AI
Responsibilities:
- Apply LLM-powered chatbots to surveys
- Use LLM technology to analyse (market) data
- Solve long-existing Market Research tasks in cutting-edge AI technologies’ context
Senior Machine Learning Scientist
Responsibilities include:
- Reward model and PPO algorithm development and model training (core of the ChatGPT model)
- Designer and main contributor of Chair, a research toolkit in Python (70% of commits), which is capable of training SFT, RM, and PPO models through one line of command
- Lead of ChatEval challenge project at DSTC 11
- Contact person of the Conversational AI project
- Contact person of DReaMS Lab work package 3
Research Intern
This was a remote internship due to COVID-19. During this period, I worked on improving the engagingness of the Replika chatbot. It was a great experience to get my hands on a chatbot system that have real users (as opposed to a common lab environment of my daily research).
Applied Science Intern
Amazon
This was a remote internship as my travel to the originally agreed location Berlin was interrupted by COVID-19 pandemic. During this period, I worked on a review summarisation task with the Subjective NLP team (based in Barcelona and Berlin).
PhD student
University of Amsterdam
- Research direction: Open-domain Dialogue Systems
- Supervisors: Maarten de Rijke, Svitlana Vakulenko
- Expecting to finish thesis writing by October 2023
Master of Engineering
Northwest A&F University
- Thesis title: Research on Feature Representation and Optimization Methods in Structured Object Tracking
- Supervisor: Jifeng Ning
Bachelor of Engineering
Northwest A&F University
- Thesis title: Implementation of Single Image Haze Removal Using Dark Channel Prior
- Advisors: Dr. Yaojun Geng and Prof. Jifeng Ning
Recent Posts
Recent Publications
Quickly discover relevant content by filtering publications.
(2023).
Weakly Supervised Turn-level Engagingness Evaluator for Dialogues.
Proceedings of the 2023 Conference on Human Information Interaction and Retrieval.
(2022).
A Simple Contrastive Learning Objective for Alleviating Neural Text Degeneration.
arXiv preprint arXiv:2205.02517.
(2020).
TLDR: Token Loss Dynamic Reweighting for Reducing Repetitive Utterance Generation.
arXiv preprint arXiv:2003.11963.
(2019).
Improving Neural Response Diversity with Frequency-Aware Cross-Entropy Loss.
The World Wide Web Conference.
(2018).
Why are Sequence-to-Sequence Models So Dull? Understanding the Low-Diversity Problem of Chatbots.
Proceedings of the 2018 EMNLP Workshop SCAI: The 2nd International Workshop on Search-Oriented Conversational AI.
Me on Twitter
Log in to view
Upcoming Events
If you’re interested in AI, Web 3, and Game Development, join me in these events around Amsterdam! Click the + button in the bottom right to subscribe to this calendar.
Contact
Drop an email below: