Software
I think software counts as research output. These are the R packages I’ve written or contributed to, most of them built to answer a question in my research. More on GitHub.
Author
kumquat
A simpler, transparent LIME for tabular data. It exposes the perturbations, the black-box predictions and the local GLM, so you can see how each explanation is built.
CRAN Explainable AI
kultarr
Anchors written in pure R. You can inspect it and debug it, it runs in parallel and caches its search, and it can draw anchors in high-dimensional tours.
In development Explainable AI
scatteR
Generates scatterplots with given scagnostic features, using simulated annealing. Presented at useR! 2022.
CRAN Data generation
sillysplines
An htmlwidget (Svelte front end) for drawing a linear-spline decision boundary and exporting it as a synthetic classification dataset.
In development Teaching
Contributor
DSjobtracker
A data package of the skills and qualifications asked for in data science job ads worldwide. I curated and cleaned the 2021 data as a consultant at the Statistical Consultancy Service, University of Sri Jayewardenepura.
CRAN Data
nic
Nature Inspired Colour palettes from Sri Lankan flora and fauna. I photographed the sources, quantised the images and built colour-blind-friendly palettes.
CRAN Visualisation
tsdataleaks
Detects data leakage between series in forecasting competitions. I sped up its core calculations with Rcpp.
CRAN Time series
Apps & tools
- Bird watcher: a Shiny and leaflet app for exploring bird-watching sites in Nigeria with GBIF data, built for the Abuja useR group.
- Netball data entry: a real-time Shiny app backed by Azure SQL for the Sri Lankan national netball team (Sports Analytics Unit, University of Sri Jayewardenepura).
- Shiny dashboards for life-sciences clients at Appsilon, plus engineering support for their R packages.