Curriculum Vitae in PDF
Please feel free to contact me at lfelipesv at gmail.com

Bio

I am currently a Postdoctoral Researcher at the Max Planck Institute for Security and Privacy (MPI-SP), Bochum, Germany.

My research aims to develop Generative AI methods for basic science applications, especially drug discovery (protein and antibody design) and climate change. For this, I am particularly interested in developing novel reinforcement learning (RL) algorithms, graph neural networks (GNNs), and diffusion-based models. Currently, I am working towards Generative AI for protein and antibody design applied to drug discovery in the following topics:

  • Generative AI for Protein/Antibody Design: protein sequence design via RL, antigen-conditioned antibody design using geometric deep learning, graph neural networks and diffusion models.
  • Protein Foundation Models that uses both 1D sequence and 3D structure information: development algorithms and training protein foundation models that incorporate 1D and 3D (multimodal) information.
  • AI-based methods for Protein/Antibody Design: training of antibody-specific models for structure prediction, inverse folding, and error prediction.

My general interests include artificial intelligence, deep reinforcement learning, and applied deep learning. From these, I am particularly interested in:

  • Deep Reinforcement Learning: improving sampling efficiency and performance by developing new credit assignment and exploration methods, goal-conditioned RL, graph structure refinement via RL.
  • Graph Neural Networks: development of novel algorithms to be applied to target-conditioned protein design, applying GNNs to error prediction.
  • Diffusion Models: application of diffusion-based models for 3D structure generation, development of diffusion algorithms for 3D full atom protein structure generation.

I have previous research experiences in Deep Reinforcement Learning applied to Robotics (continuous control, development of simulated environments, multi-agent systems), Speech Synthesis (corpus building and segmentation, statistical parametric speech synthesis), and Natural Language Processing (using sequence-to-sequence models combined with LSTMs and Transformers).

My personal interests include traveling, sports (football, kickboxing, skate, snowboard, surf, surfskate), music (classical guitar, cavaquinho) and reading (fantasy, science fiction, mountaineering).

Research Interests

  • Generative AI for Basic Science (Protein and Antibody Design, Drug Discovery, Climate Change)
  • Deep Reinforcement Learning (Credit Assignment, Exploration, Multi-agent Systems, Goal-conditioned RL)
  • Graph Neural Networks (Generative Models, Equivariance, Graph Structure Optimization)

Positions

2024.11-Present: Max Planck Institute for Security and Privacy (MPI-SP), Germany: Postdoctoral Researcher. Adviser: Meeyoung Cha
2021.10-2024.9: Data Science Group, Institute for Basic Science (IBS), South Korea: Senior Researcher. Adviser: Meeyoung Cha
2021.3-2021.9: Mechanical Engineering Research Institute, KAIST, South Korea: Postdoctoral Researcher. Adviser: Dongsoo Har

Education

2017-2021: Korea Advanced Institute of Science and Technology (KAIST): PhD degree in Green Transportation (Thesis: Performance Enhancement in Multigoal Reinforcement Learning using Hindsight Experience Replay). Adviser: Dongsoo Har
2015-2017: Federal University of Rio de Janeiro (UFRJ): Master’s Degree in Electrical Engineering. (Dissertation: Comparison between rule-based and data-driven natural language processing algorithms for Brazilian Portuguese speech synthesis). Adviser: Fernando Gil Vianna Resende Junior
2013: Korea Advanced Institute of Science and Technology (KAIST): Exchange Student at the Electrical Engineering department
2009-2015: Federal University of Rio de Janeiro (UFRJ): Bachelor’s Degree in Electronic and Computer Engineering

Internships

Summer 2013: Hyundai Motor Company, Namyang Research and Development Center: Eco Vehicle Control System Development Team

Current Projects

Development of AI-based methods for drug discovery: Research on Protein/Antibody design methods using Deep Learning for the development of new therapeutics/vaccines.

Past Projects

AI World Cup: Robot Soccer competition. Software Developer. Author and Maintainer of the Python examples and support.
Selected Media Coverage: Korea Herald, AI Times

WCG AI Masters: Design and development of the online platform.
Selected Media Coverage: Press Release, Youtube

Robot Hand Project: Research in Reinforcement Learning algorithms for Robot Control.
Selected Media Coverage: KAIST Breakthroughs 2020

EBS ESOF Platform: System modeling, curriculum design and back end development. Web platform to teach robotics and deep reinforcement learning using Webots and Google Blockly.
Online Platform: EBS Platform

Publications

Please note that the † symbol denotes a mentored student.

Peer-Reviewed International Conferences and Journals

[P1]: Vecchietti, Luiz Felipe & Lee, Minji, et al (2024). Robust Optimization in Protein Fitness Landscapes Using Reinforcement Learning in Latent Space. International Conference on Machine Learning (ICML) 2024.

[P2]: Seo†, Minah & Vecchietti, Luiz Felipe & Lee, Sangkeum & Har, Dongsoo. (2019). Rewards Prediction Based Credit Assignment for Reinforcement Learning with Sparse Binary Rewards. IEEE Access. PP. 1-1. 10.1109/ACCESS.2019.2936863.
Selected as a featured research for the KAIST Breakthroughs Magazine in 2020

[P3]: Vecchietti, Luiz Felipe & Seo, Minah & Har, Dongsoo. (2020). Sampling Rate Decay in Hindsight Experience Replay for Robot Control. IEEE Transactions on Cybernetics.

[P4]: Vecchietti, Luiz Felipe, et al. “Batch Prioritization in Multigoal Reinforcement Learning.” IEEE Access 8 (2020): 137449-137461.

[P5]: Lee, Sangkeum & Vecchietti, Luiz Felipe et al. (2020). Power Management by LSTM Network for Nanogrids. IEEE Access.

[P6]: Lee, Sangkeum & Jin, Hojun & Vecchietti, Luiz Felipe et al. Short-term predictive power management of PV-powered nanogrids. IEEE Access, v. 8, p. 147839-147857, 2020.

[P7]: Kim†, Taeyoung & Vecchietti, Luiz Felipe et al. “Machine Learning for Advanced Wireless Sensor Networks: A Review.” IEEE Sensors Journal (2020).

[P8]: Lee, Sangkeum & Jin, Hojun & Vecchietti, Luiz Felipe et al. “Power Management of Nanogrid Cluster with P2P Electricity Trading Based on Future Trends of Load Demand and PV Power Production.” arXiv preprint arXiv:2009.00863 (2020).

[P9]: Kim, Sungkwan & Kim, Inhwan & Vecchietti, Luiz Felipe et al. “Pose Estimation Utilizing a Gated Recurrent Unit Network for Visual Localization.” Applied Sciences 10.24 (2020): 8876.

[P10]: Lee, Sangkeum & Har, Dongsoo & Vecchietti, Luiz Felipe et al. “Optimal Link Scheduling Based on Attributes of Nodes in 6TiSCH Wireless Networks.” 한국정보기술학회논문지 18.1 (2020): 77-92.

[P11]: Hong, Chansol & Jeong, Inbae & Vecchietti, Luiz Felipe et al. “AI World Cup: Robot Soccer-Based Competitions.” IEEE Transactions on Games (2021).

[P12]: Kim†, Taeyoung & Vecchietti, Luiz Felipe et al. “Two-stage training algorithm for AI robot soccer.” PeerJ Computer Science (2021).

[P13]: Lee, Sangkeum & Jin, Hojun & Vecchietti, Luiz Felipe et al. “Cooperative decentralized peer-to-peer electricity trading of nanogrid clusters based on predictions of load demand and PV power generation using a gated recurrent unit model.” IET Renewable Power Generation, vol. 15, pp. 3505-3523 (2021).

[P14]: Mishra, Sumit, & Rajeendran, Praveen Kumar & Vecchietti, Luiz Felipe et al. “Sensing accident-prone features in urban scenes for proactive driving and accident prevention”, IEEE Transactions on Intelligence Transportation Systems (2023).

Workshops, Domestic Conferences, Posters, and Extended Abstracts

[W1]: Wijaya†, Bryan Nathanael & Vecchietti, Luiz Felipe et al. “Evaluation of Antibody Structure Reconstruction with an SE(3)-Equivariant Graph Neural Network.” presented at the Korea Software Congress (KSC), Busan, South Korea (2023).

[W2]: Jung†, Hyunkyu & Vecchietti, Luiz Felipe et al. “Protein Structure Tokenizer for Efficient Learning.” presented as a poster in Peptalk, San Diego, USA (2023).

[W3]: Hangeldiyev†, Begench & Rzayev, Anar & Armanuly, Azamat & Jung, Hyunkyu & Vecchietti, Luiz Felipe et al. “Antibody Sequence Design With Graph-Based Deep Learning Methods.” presented at the Korea Software Congress (KSC), (2022).

[W4]: Lee1†, Minji & Vecchietti1, Luiz Felipe et al. “Protein Sequence Design in a Latent Space via Model-based Reinforcement Learning.” presented at the NeurIPS Machine Learning in Structural Biology (MLSB) Workshop, (2022).
1 denotes co-first authors.

[W5]: Lee†, Minji & Rzayev, Anar & Jung, Hyunkyu & Vecchietti, Luiz Felipe et al. “Structure-based representation for protein functionality prediction using machine learning.” presented at the Korea Computer Congress (KCC), (2022).

[W6]: Rajendran†, Praveen Kumar & Mishra, Sumit & Vecchietti, Luiz Felipe et al. “RelMobNet: End-to-end relative camera pose estimation using a robust two-stage training.” presented at the ECCV IWDSC Workshop, (2022).

Academic Services

Reviewer

Journals: IEEE Transactions on Cybernetics, IEEE Transactions on Games, IEEE Sensors, Frontiers in Robotics and AI

Conferences: AAAI ICWSM 2022, NeurIPS 2024

Workshops: ICML LatinX Workshop 2021, ICLR Reincarnating RL Workshop 2023, NeurIPS Machine Learning in Structural Biology (MLSB) Workshop 2023, ICML ML4LMS Workshop 2024

Invited Talks

[I1]: Developing and applying deep learning methods for protein design
Graduate School of AI, Gwangju Institute of Science and Technology (GIST), July 2023.

[I2]: Developing and applying deep learning methods to facilitate new scientific discoveries
Max Planck Institute for Security and Privacy (MPI-SP), May 2023.

[I3]: Target-conditioned Protein and Antibody Design for Drug Discovery
IBS Winter School on AI-Boosted Basic Science, Institute for Basic Science, Dec 2022.
Co-delivered with Prof. Ho Min Kim (KAIST)

[I4]: Identifying the key actions that lead an agent to accomplish a task in model-based reinforcement learning
School of AI Convergence, Chonnam National University, Nov 2021.

[I5]: Performance enhancement in multigoal model-based deep reinforcement learning
Cho Chun Shik Graduate School of Mobility, KAIST, Oct 2021.

[I6]: Identifying the key actions that lead an agent to accomplish a task in model-based reinforcement learning
Data Science Group, Institute for Basic Science, Apr 2021.

Mentored Students

It is a great pleasure to work with very talented and hardworking students.

I am very thankful for the support and trust given from my advisors throughout my career: Prof. Fernando Gil Vianna Resende Junior (MsC), Prof. Dongsoo Har (PhD), and currently Prof. Ho Min Kim, and Prof. Meeyoung Cha.

Current

  • Bryan Nathanael Wijaya, 2023-Now
    MsC Candidate at KAIST
    Topic: Protein Sidechain Reconstruction using Deep Learning, Antigen-conditioned Antibody Design

  • Begench Hangeldiyev, 2022-Now
    Chemical Engineering / CS Undergraduate at KAIST
    Topic: Antibody Sequence Design with Graph-Based Methods
    Published his first paper on antibody design with deep learning at KSC 2023.

  • Hyunkyu Jung, 2021-Now
    PhD Candidate at KAIST
    Topic: Protein-Protein Interactions using Equivariant Point Cloud Transformers Submitted a first-author paper on protein-protein interactions using deep learning to ICCV 2023.

  • Anar Rzayev, 2022-Now
    CS Undergraduate at KAIST
    Topic: Antibody Structure Representation Learning

Alumni

  • Sangmin Lee, 2023 CS Undergraduate at KAIST
    Topic: AI-based methods for design thermostable proteins

  • Maxim Krassimirov Mintchev, 2022 (now Research Associate at TU Berlin)
    Mechanical Engineering Master’s Candidate at Karlsruhe Institute of Technology and KAIST (Double Degree)
    Topic: Recommendation Systems applied to Logistics Applications
    Worked and successfully defended his master’s thesis on recommendations systems at KIT and KAIST.

  • Minji Lee, 2022
    CS Undergraduate at KAIST
    Topic: Protein Optimization using Deep Reinforcement Learning
    Submitted a first-author paper on protein engineering to ICML 2024. Published two first-author papers on protein functionality prediction and protein engineering using deep learning at KCC and NeurIPS MLSB Workshop 2022.

  • Lucas Santiago Peixoto0, 2022
    Computer Engineering Undergraduate at Federal University of Rio de Janeiro (UFRJ)
    Topic: Sentence-level Text Analysis using Transformers
    Worked and successfully defended his undergraduate thesis on NLP.
    0Co-advisor in undergraduate thesis with Prof. Fernando Gil Vianna Resende Junior

  • Matheus Tymburiba Elian, 2021-2022 (now Faculty at the Federal University of Minas Gerais)
    Industrial Design PhD Candidate at University of Tsukuba
    Topic: Gender-Ambiguous Voice Agents for Voice User Interfaces
    Submitted a first-author paper on gender-ambiguous voice agents to CHI 2023.

  • Kien Hoang, 2021-2022 (now MsC Candidate at EPFL)
    Mathematics Undergraduate at KAIST
    Topic: Graph structure optimization with Deep Reinforcement Learning

  • Sumit Mishra, 2020-2022 (now PhD Candidate at KAIST)
    Robotics Master’s Candidate at KAIST
    Topic: Intelligent Transportation Systems using Deep Learning
    Submitted a first-author paper on ITS to IEEE Transactions on Intelligent Transportation Systems.

  • Praveen Kumar Rajendran, 2020-2022 (now Computer Vision Engineer at Neubility) Future Vehicle Program Master’s Candidate at KAIST
    Topic: Camera Pose Estimation using End-to-End Deep Learning
    Published a first-author paper on pose estimation with deep learning at ECCV IWDSC Workshop 2022. Submitted a first author paper to CVPR 2023.

  • Taeyoung Kim, 2019-2021 (now PhD Candidate at KAIST)
    Green Transportation Master’s Candidate at KAIST
    Topic: Deep Reinforcement Learning, Cooperative-Competitive Multi-agent Systems
    Published a first-author review paper on sensor networks using deep learning at IEEE Sensors Journal. Published a first-author paper on deep reinforcement learning in multi-agent environments at PeerJ Computer Science.

  • Kyujin Choi, 2019-2020 (now ML Engineer at KT)
    Green Transportation Master’s Candidate at KAIST
    Topic: Deep Reinforcement Learning, Cooperative-Competitive Multi-agent Systems
    Published a co-first author paper on deep reinforcement learning in multi-agent environments at PeerJ Computer Science.

  • Minah Seo, 2018-2019 (now ML Engineer at KT)
    Green Transportation Master’s Candidate at KAIST
    Topic: Credit Assignment in Deep Reinforcement Learning
    Published a first-author paper on efficient credit assignment in RL at IEEE Access.