Project documentation

Technical docs for the AI, data science, and knowledge graph projects I build and maintain. Each site covers architecture decisions, data models, deployment notes, and lessons learned.


ML pipeline time series AWS Lambda

Fitness dashboard

14 years of workout data, an ML classifier that separates runs from dog walks, and a full ETL pipeline from CSV export to interactive dashboard.

Multi-LLM Conversation AI Streamlit

ConvoScope

A unified interface for comparing outputs across OpenAI, Anthropic, and Google models — with persistent history and automatic topic extraction.

Computer vision Knowledge graph Neo4j

Beehive analytics

Computer vision and metadata extraction turn four years of hive inspection photos into a queryable knowledge graph with weather correlation.

NLP Timeline extraction GCP

ChronoScope

Extracts temporal events from unstructured text and visualizes them as interactive timelines — surfacing narrative structure from documents.

Knowledge graph Wikidata Neo4j

Digital memory chest

Transforms family stories and memories into a 14-entity knowledge graph enriched with Wikidata, preserving relationships and provenance across generations.

RAG Evaluation harness FastAPI

Poolula platform

AI-powered LLC management combining structured data, document search via ChromaDB, and a multi-provider RAG chatbot with a built-in evaluation harness.

TransE ML Neo4j + Streamlit

Citation compass

Predicts which academic papers are likely to cite each other using a from-scratch TransE implementation over knowledge graph embeddings, with interactive network exploration.