Experiments, prototypes, and research notes.

Smart Document Assistant

An experiment exploring how AI can help organize and summarize complex documents while keeping a human reviewer in the loop.

Problem:

Long, inconsistent documents make specific details slow to find.

Approach:

Document extraction, structured summarization, and human review.

Data:

Illustrative

Human review required

View details

Plain Language Rewriter

Testing whether a model can rewrite dense administrative text into clearer language without changing its meaning.

Problem:

Official wording is often accurate but difficult for the intended reader.

Approach:

Constrained rewriting with a verification pass.

Data:

Synthetic

Human review required

View details

Repetitive Task Mapper

A small automation study that maps where repetitive manual steps appear in a workflow before any code is written.

Problem:

Automation projects often start with tooling instead of with the actual bottleneck.

Approach:

Clustering of task descriptions with an LLM summary of each cluster.

Data:

Synthetic

Human review required

View details

Semantic Project Search

Comparing keyword search with embedding-based search across a small case-study archive.

Problem:

Keyword search fails when the visitor and the author use different vocabulary.

Approach:

Text embeddings with cosine similarity ranking.

Data:

Illustrative

View details