// projects
A collection of personal projects across data science and AI. Each one is published here as it is ready, so do check back from time to time.
Profiles any dataset and writes out the story in it, with every method left visible and tweakable.
Runs the right statistical test, checks its assumptions first, and shows the arithmetic.
Publication-grade charts across eighteen families, each one handing you the code that drew it.
Cleans and engineers features with real pandas in the browser, recording every step as a recipe you can replay.
Point it at a table and get a leaderboard of tuned models, each one openable down to the tree.
Decomposition, backtesting and forecasts with honest intervals, classical models and neural ones scored side by side.
Topics, sentiment, summarization and search on raw text, with the lexicon and the neural model free to disagree.
An open evaluation engine for LLM and ML systems - metrics, fairness, and drift, all inspectable.
Read papers with the author's own explanations layered over them, in a portable file you can pass on.
A notebook-style SQL explorer over local and remote sources.
End-to-end analysis for Likert and categorical survey data.
Mapping and spatial analysis for location-tagged datasets.
Privacy-preserving synthetic tables that keep the statistics, drop the identities.
Image classification and detection demos with explainable overlays.
Collaborative and content-based recommenders with cold-start handling.
Statistical and model-based outlier detection for streams and tables.
Estimate treatment effects from observational data, honestly.
Retrieval and grounding quality checks for RAG pipelines.
Bias and parity diagnostics across protected segments for ML models.
Drift, degradation, and stability monitoring for models in production.
Lightweight tracking for runs, params, and metrics across projects.
Templates and scripts for packaging, serving, and versioning models.
TM