Portfolio

Projects

A selection of my work, each with a case study. Many more projects are on my GitHub.

• pokémonpokémonpokémonpokémonYOLOv11s

  pokémonpokémonpokémonpokémonRT-DETR

  pokémonpokémonpokémonpokémonFaster R-CNN

  9 classes, balanced

  yolov11s vs rt-detr vs faster r-cnn

  01Computer vision

  Pokémon detection, three ways

  YOLOv11s, RT-DETR and Faster R-CNN trained on the same nine Pokémon classes and compared on speed, classification and box quality.

  per frame with YOLOv11s, about 8× faster than Faster R-CNN

  11 ms

  F1 for YOLOv11s, the best of the three

  0.89

  mAP 50-95 for Faster R-CNN, the tightest boxes

  0.80

  • PyTorch
  • Ultralytics
  • YOLOv11s
  • RT-DETR
  • Faster R-CNN
  • OpenCV

  Case studyRead the case studyCode
• 90 steps observed30 forecast4 of 76 variables · Seq2Seq GRU

  seq2seq gru · 90 steps in, 30 out

  02Time series

  Weather, 30 steps ahead

  A Seq2Seq GRU in PyTorch that reads 90 steps of history and forecasts 76 weather variables 30 steps ahead, for 422 stations. Built for the Kaggle challenge of a deep learning course.

  weather variables forecast by one model

  76

  stations, 192,432 training sequences

  422

  best validation MAE over 30 steps

  0.557

  • PyTorch
  • GRU
  • Seq2Seq
  • Scikit-Learn
  • Pandas

  Case studyRead the case studyCode
• Projected gradient · Frank-Wolfe5-clique

  projected gradient descent · frank-wolfe

  03Optimization

  Max clique, made continuous

  The maximum clique problem recast as continuous optimization over the simplex, and solved with projected gradient descent and three variants of Frank-Wolfe. Built for the Optimization for Data Science course in Padova.

  execution time on the DIMACS graphs

  −30%

  found the largest cliques on average

  PGD + L0

  moved the least when the step size changed

  FW, AFW

  • Python
  • NumPy
  • SciPy
  • Frank-Wolfe
  • Projected Gradient Descent
  • DIMACS

  Case studyRead the case studyCode