Projects

Code from my research, data-science training and AI tooling. Every project is open on GitHub.

AI tools for researchers

hallucination-check · Claude skill · 2026

↳ Checks whether an AI answer can be trusted. It pulls out the facts that can actually be tested (numbers, citations, names, computed results), checks them against your own data first and only then runs one targeted search, and reports a hallucination score with just the claims that failed. It caps itself at eight claims so the check stays cheap.

analysis-brief · Claude skill and plugin · 2026

↳ Turns pasted tables, logs and long chat threads into a one-page brief: the data reduced to its schema, plus the decisions, failed attempts and finish line. Paste the brief into a fresh chat and carry on without re-sending the data. In testing, a new chat given only the brief wrote a correct, working R figure script.

Machine learning in R

SVM recursive feature elimination · R · caret · ggplot2 · 2026

↳ Finds the smallest set of variables that still predicts well. A linear SVM ranks features, drops the weakest and retrains, with 5-fold cross-validation at each step; the pipeline plots accuracy against the number of features and ships with a synthetic dataset for testing.

SVM feature ablation · R · caret · ggplot2 · 2026

↳ The reverse test: it removes the most important features in steps and retrains, measuring how fast accuracy falls on unseen data. This shows which variables a model truly depends on and how robust it is.

Deep learning and data science in Python

Plant seedling classification · Python · TensorFlow/Keras · OpenCV

↳ A convolutional neural network that identifies 12 plant-seedling species from images, a building block for automated weed and crop monitoring. With data augmentation and learning-rate scheduling the final model reached about 80% accuracy on held-out images.

FoodHub business insights · Python · pandas · seaborn

↳ Exploratory analysis of a food-delivery order dataset. Over 71% of orders come at weekends, most customers order only once, and faster deliveries go with slightly better ratings, which led to recommendations on retention and weekday promotions.

Bioinformatics

iDEP (fork) · R · Shiny · RNA-seq

↳ My fork of iDEP, the open-source platform from South Dakota State University for RNA-seq differential expression and pathway analysis, which I keep as my RNA-seq analysis workflow.

The two Python projects come from the Post Graduate Program in AI and Machine Learning (UT Austin with Great Learning).