BIO

I am a Research Scientist at Google DeepMind. I contributed to several large-scale projects such as AlphaStar, AlphaCode, and Veo. Before joining Google DeepMind, I received my PhD in machine learning from the MILA lab at the University of Montreal in 2018, where I was fortunate to be advised by Professor Yoshua Bengio. 

Research Interest

I have a broad interest in various topics in deep learning, from deep RL (AlphaStar), LLMs (AlphaCode) to large-scale diffusion models (Veo).

• Language modeling (PhD thesis)

• RL (AlphaStar)

• Code generation (AlphaCode)

• Video generation (Veo)

Here is a link to my CV

News

• Veo was shipped at Google I/O 2024.

• AlphaCode paper has been published in Science [webpage], and it was highlighted as one of the breakthroughs of the year [webpage].

• I am thrilled to share our new paper about AlphaCode [blog][pdf]

• Our paper, Step-unrolled Denoising Autoencoders for Text Generation, got accepted to ICLR 2022 [pdf]

•  AlphaStar paper has been published in Nature [webpage]

• My PhD dissertation became publicly accessible [pdf]

• The AlphaStar team has released the first result on StarCraft II [DeepMind blog]

• I joined DeepMind as a research scientist.

• Our paper, Hierarchical Multiscale Recurrent Neural Networks, got accepted to ICLR 2017 [pdf]

• I did a fall Internship at Google Research, Mountain View, 9/2016 - 12/2016

• Our paper, A Character-Level Decoder without Explicit Segmentation for Neural Machine Translation, got accepted to ACL 2016 (oral) [pdf][code]

• I was a visiting PhD student at NYU machine learning group (CILVR, working with Professor Kyunghyun Cho), 10/2015 - 4/2016

• Our paper, Detecting Interrogative Utterances with Recurrent Neural Networks, got accepted as the best workshop paper at NeurIPS 2015 Workshop on Machine Learning for Spoken Language Understanding and Interactions

• Our paper, A Recurrent Latent Variable Model for Sequential Data, got accepted to NeurIPS 2015 [pdf][code]

• I did a summer internship at Microsoft Research, Redmond, 6/2015 - 9/2015

• Our paper, Gated Feedback Recurrent Neural Networks, got accepted to ICML 2015 [pdf]

Talks

• AlphaCode

  • Google Speech, virtual, 4/2022

  • Google Translate, virtual, 3/2022

  • HKUST, virtual, 3/2022

  • Naver, Pangyo, 12/2022

  • Yonsei University, Seoul, 12/2022

  • Korea University, Seoul, 12/2022

• AlphaStar

  • Kakao Brain, virtual, 6/2021

  • UNIST, virtual, 5/2021

  • Hanyang University, Seoul, 8/2019

• Neural Network Architectures for Time-Series and Machine Learning Problems, AI Korea Summer School @Yonsei University, Seoul, 7/2019

• Towards Hierarchical Multiscale Recurrent Neural Networks

  • Toyota Institute of Technology at Chicago, Chicago, 6/2017

  • Naver, Pangyo, 9/2017

  • KakaoBrain, Pangyo, 9/2017

• Hierarchical Multiscale Recurrent Neural Networks @Google Research, Mountain View, 10/2016

• A Character-Level Decoder without Explicit Segmentation for Neural Machine Translation @ACL'16, Berlin, 8/2016

• A Character-Level Decoder for Neural Machine Translation and Multiscale Recurrent Neural Networks @DeepMind, London, 8/2016

• Deep Recurrent Neural Networks for Sequence Modelling

  • Samsung Advanced Institute of Technology, Suwon, 4/2016

  • Hyundai Motors, Uiwang, 4/2016

• Detecting Interrogative Utterances with Recurrent Neural Networks @NeurIPS Workshop on SLUI, Montreal, 12/2015

• Variational Recurrent Neural Networks @New York University, New York City, 11/2015

• Gated Feedback Recurrent Neural Networks @ICML'15, Lille, 7/2015