Kelve Cartaxo

CS student building machine learning systems. Heading into generative AI.

  • Federal University of Ceará, Brazil, class of 2027
  • open to remote workget in touch

about

Researcher at NEMO, the machine learning and optimization lab at UFC, Brazil.

Computer science, graduating 2027.

Python-first: I train models, explain them and ship them behind an API.

Next goal: doing research in Japan.

based in
Russas, Brazil (UTC−3)
focus
ML engineering, generative AI
status
open to remote work
english
C2 (EF SET 72/100)
japanese
basic, N5 level (studying)
Kelve Cartaxo

now

  • thesisF1 podium prediction, XGBoost vs LSTM
  • researchML for the maximum diversity problem
  • buildingNEMO lab server distro
  • learningMLOps, PostgreSQL

stack

  • Python
  • PyTorch
  • XGBoost
  • SHAP
  • Hugging Face
  • YOLOv8
  • ONNX
  • FastAPI
  • Docker
  • SQLite
  • Git
  • LaTeX

projects

  • deployed 2026

    UFC Russas campus assistant

    Chatbot that answers from the campus website, not from model memory. Runs on 2 GB, no GPU.

    • RAG
    • FastAPI
    • ChromaDB
    • Gemini
  • thesis 2026

    F1 podium prediction

    252 races, 2014 to 2025. Can ML beat the starting grid? The gain shows up in ranking.

    • XGBoost
    • LSTM
    • SHAP
  • research 2026

    Maximum diversity, with ML

    Picking the items that are farthest apart from each other. Live demo in the lab below.

    • graphs
    • heuristics
    • Python
  • prototype 2026

    NEMO lab server

    A server distro so lab members can host anything without knowing Linux.

    • Ubuntu
    • Docker
    • PostgreSQL
  • coursework 2026

    Turing machine simulator

    Theory of Computation project, presented in a seminar. Pair project.

  • coursework 2026

    Compiler

    Compilers course project. Pair project.

  • planned next

    Steam platinum helper

    Next build: a helper for finishing every achievement in a game.

    • FastAPI
    • Docker
    • PostgreSQL

max diversity, live

Pick 8 of 40 points as far apart as possible. Local search, running now.

total distance 0.000
step 0

toy instance, not research results

games

Three algorithms to play against. Try to beat them.