Projects

Selected work

Each entry records the problem, the decisions that mattered, and the measurable outcome. Filter by discipline below.

AI Training2026

AI agents that answer customer reviews

Led end-to-end development of review-response agents on OpenAI and Claude, with tool integrations and memory for autonomous operation — and a prompt framework that cut model cost 18%.

18% reduction in model cost with no loss of response quality
Tool integrations and memory for autonomous, multi-step operation

PythonNode.jsOpenAIClaude

AI Training2025

Retrieval that stopped the model inventing facts

A RAG layer over each business's own details, so generated responses referenced things that were actually true — and said nothing when the source material didn't cover it.

Business facts retrieved per response instead of stuffed into every prompt
Retrieval scoped per location, so no account could cite another's details

PythonVector databasesRAGOpenAI

AI Training2023

Sentiment analysis that survived contact with real reviews

A pipeline turning raw review text from several platforms into reputation insights — and the calibration work needed before anyone would act on the numbers.

Multi-platform review text normalised into one analysable shape
Model output calibrated against human labels before it was trusted

PythonSQLNoSQLData pipelines