Profile photo

Kleyton da Costa

AI Researcher · Incoming Ph.D. @ UCL

kleyton.vsc@gmail.com · kleyton.costa@holisticai.com

I'm an AI researcher working at the intersection of AI safety, graph learning, optimization, and probabilistic models. In October 2026 I'll begin a Ph.D. in Computer Science at University College London, advised by Prof. Philip Treleaven and Prof. Dimitrios Kanoulas, where my research will focus on predictive safety in embodied AI , methods for anticipating unsafe behavior in physical and decision-making agents before it occurs.

Recent work includes a 2026 survey of predictive safety in embodied AI with collaborators at UCL and Holistic AI, and the development of divergence-guided particle swarm optimization for robust search.

I'm currently an AI Researcher and Engineer at Holistic AI and a core contributor to open-source research software for AI trustworthiness, optimization, and state-space modeling.

Recent Publications

All publications →

  1. A Survey on Predictive Safety in Embodied AI

    2026

    Kleyton da Costa, Adriano Koshiyama, Dimitrios Kanoulas, Philip Treleaven

    SSRN

  2. Divergence-Guided Particle Swarm Optimization

    2026

    Kleyton da Costa, Bernardo Modenesi, Ivan F. M. Menezes, Hélio Lopes

    arXiv:2604.12001

  3. The Confidence Manifold: Geometric Structure of Correctness Representations in Language Models

    2026

    Seonglae Cho, Zekun Wu, Kleyton Da Costa, Adriano Koshiyama

    arXiv:2602.08159

  4. Time series forecasting via integrating a filtering method: an application to electricity consumption

    2025

    Felipe Leite Coelho da Silva, Josiane da Silva Cordeiro, Kleyton da Costa, Nemias Saboya, Paulo Canas Rodrigues, Javier Linkolk Lopez-Gonzales

    Computational Statistics

  5. Learning on Graphs via Generalized Divergence Measures

    2025

    Kleyton Vieira Sales da Costa

    PUC-Rio

Research Software

holisticai

Open-source library for assessing AI trustworthiness

optymus

Optimization methods for science and engineering

compute-geometry

Research-focused computational geometry library for Python

dynaris

JAX-powered library for state-space models

Education

  • Ph.D. in Computer Science · University College London
    Oct 2026 — Oct 2030 (incoming)
  • M.Sc. in Computer Science · PUC-Rio
    2023 — 2025
  • B.Sc. in Economics · UFRRJ
    2017 — 2021

Affiliations

  • AI Engineer · Holistic AI
    Aug 2025 — Present
  • AI Researcher · Holistic AI
    Oct 2021 — Present
  • Researcher · PUC-Rio
    Mar 2023 — Aug 2025
  • Data Scientist · State Government of Rio de Janeiro
    Jan 2022 — Mar 2023

Contact

For research collaboration, academic discussion, or open-source contributions, reach me at kleyton.vsc@gmail.com.

Kleyton da Costa