Join Us!
How to join?
If you are interested in joining our lab, please tell me a bit about yourself using this form. It's very important that you submit this form to be considered for joining our lab. Due to the high volume of applications, it's extremely difficult to respond to all applicants. Please understand that I won't answer any application emails that are not submitted through this form.
Frequently Asked Questions
- Q: [📢 Updated on 2026.07.17] How many students are you planning to take on in 2027?
- A: I am looking to take on up to two new students in 2027.
- Q: How important is data in your research?
- A: Very. You should be data-pilled. Better models don't always come from better algorithms; often, they come from better data. I expect you to think seriously about what data we train on, how we collect or synthesize it, how we select and evaluate it, and ultimately what the model can learn from it. If you join our lab, you should care deeply about data, not treat it as something that happens to be fed into a model.
- Q: What does being "data-pilled" mean? 💊
- A: It is a play on the red-pill/blue-pill metaphor from The Matrix (1999): taking the red pill means waking up to a deeper reality. To be data-pilled is to accept a simple but consequential idea: data is one of the fundamental sources of intelligence. Some people often talk about intelligence as if it comes primarily from learning algorithms, objectives, or architectures. But a model can only learn from the experiences we give it. Its knowledge, reasoning patterns, preferences, blind spots, and even many of its capabilities are shaped by its data. Once you become data-pilled, you stop treating data as a passive ingredient that gets fed into a model. You start asking a different question: what experiences would produce the intelligence we want? That shift changes how you think about AI research. Not data-pilled yet? We'll get you there.
- Q: What are other expectations from prospective students?
- A: I expect you to be self-motivated and passionate about research. You should be open to learning new things and able to adapt quickly in the fast-evolving field of AI. (I know—it just keeps getting faster and faster, right? 😅)
You should also enjoy collaborating with others and actively participating in research discussions within the lab.
In terms of coursework, you should have taken several AI-related classes (e.g., ML, NLP, CV, RL). As for technical skills, you should be comfortable using Python and PyTorch, and familiar with at least one training framework (e.g., miles, verl, trl, llama factory, unsloth).
- Q: Do I need to work with your lab first before I can apply?
- A: While not mandatory, I strongly encourage you to have a hands-on experience (e.g., internship) with our lab first before officially applying through KAIST AI. Applicants who have prior experience with our lab will be given higher priority during the admission process.
- Q: What will I be working on during my internship?
- A: You will work closely with one of our graduate students (or with me) on a pre-defined research project. If the work leads to meaningful results, we will aim to submit it to a top-tier conference, and you could be invited to join as a co-author if your contributions are significant.
- Q: Do I have to speak English?
- A: We will be using English as the primary language for written communication and research presentations. Spoken communication can be in Korean if you prefer, but we do provide an English-friendly environment. So please be prepared to use English. I completely understand that this can be challenging for those whose first language is not English (it is for me too 😂). But it's actually very helpful for you to get used to using English for research, earlier the better. It allows you to build confidence when presenting your work at conferences and collaborating with international researchers (e.g., NVIDIA). Trust me, it will pay off tremendously in the long run!
