Projects
Open-source contributions and personal projects, mostly in R.
Data & ML engineering
- tidypredict — XGBoost → SQL
- Contributed the translation of XGBoost models into SQL, so trained models can be scored inside the database without an R or Python runtime.
- treesnip
- LightGBM and CatBoost engines for the tidymodels framework.
- embed —
step_woe() - Contributed a weight-of-evidence preprocessing step, used in credit-risk feature engineering.
Data ingestion & APIs
- wikiaves
- R client for the WikiAves API — collects bird observation data and audio metadata.
- wavesurfer
- R/Shiny interface to wavesurfer.js for audio visualisation and labelling.
Research
- mestrado — acoustic bird detection
- M.Sc. project (Polytechnic School, University of São Paulo) on automatic detection of nocturnal bird species from field recordings. An R package that reproduces the pipeline: collecting recordings (xeno-canto, WikiAves), slicing and labelling audio, train/test splits, and a 1D convolutional network on raw audio.
Deep learning
- torchaudio
- Audio deep-learning tools for R.
- facenet
- Face-recognition implementation with Torch.
- torchdatasets
- Contributed a dataset to the R torch datasets package.
More on GitHub ↗