About me

Davide 刘大为

Education:

  • 2019 – 2021 Master degree in Computer Science at Tsinghua University, Beijing, China.
  • 2016 – 2019 Bachelor degree in Computer Science at Padova University, Padova, Italy.

Languages: Italian, English, Chinese.

Interests: Deep Learning, Computer vision, Algorithms, Matematichs, Videogames, Japanese anime, Chinese language, Track and Field, Marathon, OCR.



Short bio

Davide Liu completed his Bachelor’s in Computer Science in Italy at the University of Padova in 2019 and obtained his Master’s in Advanced Computing at Tsinghua University in 2021 conducting research in Reinforcement Learning and Security in AI. His goal is to use AI in a safe and smart way to improve the quality of our life. During his studies, he had the opportunity to attend internships in two AI startups: one in Italy where he improved an existing computer vision algorithm that lead to winning a challenge organized by the US Pentagon, and one in China doing research on novel architectures for recommendation systems which lead to two conference publications. Currently, he is working full-time as AI researcher at SenseTime in Beijing. Outside of his professional life, Davide is an online blogger, athlete, and videogames developer.

Academic experiences

  • 2016-2019 Padova University, Italy, Padova

    Bachelor degree in Computer Science (link).

    • SpritzSecurity & Privacy Research group.
    • VIMPVisual Intelligence and Machine Perception group.
  • 2019-2021 Tsinghua University, China, Beijing

    Master degree in Computer Science (link).

    • TSAILTsinghua Statistical Artificial Intelligence & Learning.

Professional experiences

    • Developed an algorithm to enhance the metrics result obtained performing object detection on high resolution satellite images from the xView dataset.
    • Developed and tested an algorithm to detect and track specific objects on videos exploiting Faster-RCNN and Kalman filter.
    • Used Amazon Personalize AutoML tool to build and test a banking products recommendation system.
    • Used Transformer to implement a novel banking product recommendation system performing better than state-of-the-art existing models.
    • Early development of the platform EZ-loan used to estimate credit score of unbanked customers.
    • Clustering of financial transactions with ML.
    • First place at BSF Hackathon by Banque Saudi Fransi.
    • First place at MINT Hackathon by EGbank.
    • Teaching assistant of Italian language course during the fall semester of 2020 and spring semester of 2021 (80+ students each semester).
    • Main developer of DI-engine and its ecosystem: a generalized, distributed, and scalable decision intelligence engine that supports various deep reinforcement learning algorithms.
    • Research, development, optimization, analysis, benchmark, and documentation of several deep reinforcement learning algorithms.
    • Researcher at OpenDIlLab

Professional talks

  • 2021

    SIGIR ’21Transformer-based Banking Products Recommender System (link) – Virtual Event, Canada

Conference publications

  • 2022

    IJCNN ’22BRec the Bank: Context-aware Self-attentive Encoder for Banking Products Recommendation – Oral presentation – Italy, Padova

    ICAIF ’22Sequential Banking Products Recommendation and User Profiling in One Go – US, New York

    Sport experiences

    Contact me