Jun Chan
Member Of Technical Staff at Anthropic
United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolJun Chan is a research-focused software engineer with 10 years of experience building and evaluating large-scale AI systems, currently a Member of Technical Staff at Anthropic. He has driven model-evaluation tooling and preparedness work at OpenAI, contributing notable features to the widely used openai/evals framework—introducing a Solvers abstraction, self-prompting evals, and several new solvers that improve comparative LLM analysis. His background spans both academic research (UC Berkeley, NYU) and applied R&D in autonomous vehicles, giving him a strong grounding in experimental rigor and real-world deployment. Based in the UK, he works primarily on AI safety and model evaluation and prefers email contact over LinkedIn for substantive conversations. An under-the-radar strength is his ability to bridge prompt engineering, logging/metrics, and systems-level refactors to make evaluation frameworks both more powerful and more practical.
10 years of coding experience
5 years of employment as a software developer
Master's degree, Electrical and Electronic Engineering (MEng), 1st Class Honours, Master's degree, Electrical and Electronic Engineering (MEng), 1st Class Honours at Imperial College LondonA-Levels, Mathematics, Physics, Psychology, Economics, A*, A*, A*, B, A-Levels, Mathematics, Physics, Psychology, Economics, A*, A*, A*, B at Methodist College Kuala Lumpur
English, Chinese, Malay
Github Skills (20)
openai-api10
python10
evaluation10
machine-learning10
lm10
large-language-models10
llm10
prompt-engineering10
eval10
apidoc8
api8
restful-api7
api-design7
json7
api-rest7
Programming languages (6)
Github contributions (5)
openai/evals
Sep 2023 - May 2024
Evals is a framework for evaluating LLMs and LLM systems, and an open-source registry of benchmarks.
Role in this project:
ML Engineer
ML EngineerContributions:39 reviews, 35 PRs, 34 pushes in 8 months
Contributions summary:Jun contributed significantly to the `evals` repository, primarily focused on enhancing the framework for evaluating Large Language Models (LLMs). Their work involved introducing a new "Solvers" abstraction to facilitate the comparison of different model scaffolding approaches and writing a self-prompting eval, demonstrating a deep understanding of prompt engineering and model evaluation. The user also added improvements for logging model and usage statistics and suppressed excessive logs from the OpenAI library. Further contributions included implementing a new solver, `OpenAIAssistantsSolver`, and refactoring existing solver code, as well as adding new features like few shot and self-consistency solvers, all of which directly contribute to the evaluation and comparison of different LLM models.
JunShern/MetaICL
Jan 2022 - Jul 2022
An original implementation of "MetaICL Learning to Learn In Context" by Sewon Min, Mike Lewis, Luke Zettlemoyer and Hannaneh Hajishirzi
Contributions:120 commits, 2 PRs, 48 pushes in 5 months
lukedeep-learningin-contextlewislearning-to-learn
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.