Research platforms and tools where I prove out production search ideas.

Recipe Saint and Room Saint are self-directed research platforms I designed, built, and operate end to end. They exist so I can test AI search architecture, hybrid retrieval, RAG, and assistant integrations, under production conditions before recommending them to anyone else. Alongside them are open source tools from years of search and infrastructure work; if something piques your interest, the code is there to explore.

  • Recipe Saint

    A production AI search platform I built end to end: hybrid lexical and vector retrieval over more than a million recipes, RAG pipelines for query understanding and result synthesis, menu planning and shopping lists, and a Model Context Protocol server so AI assistants can search it directly. An iOS client is in development.

    recipesaint.com

  • Room Saint

    A hotel discovery platform covering roughly a million properties worldwide, built on the same research architecture as Recipe Saint: hybrid retrieval, geographic search, community ratings, and trip planning, exposed to AI assistants through a Model Context Protocol integration.

    roomsaint.com

  • Scrapy Site to Disk

    The front half of a search ingestion pipeline: a Scrapy spider that crawls a site and stores every page as raw HTML plus a structured JSON document extracted with Newspaper3k, shaped for bulk loading into Solr, OpenSearch, or Elasticsearch.

    github.com

  • AWS Solr Installer

    Infrastructure as code that stands up a production-ready Apache Solr server on AWS in minutes: Terraform provisions the EC2 instance, Ansible installs Solr with checksum verification and registers it as a systemd service.

    github.com

  • Site24x7 Solr Plugin

    A custom monitoring plugin that fills a gap in Site24x7: it polls the Solr cores STATUS API and reports per-core health and document counts, giving you dashboards and alerts for the earliest signs of indexing problems.

    github.com

  • Common Crawl Iterator

    A multithreaded Python utility for working through Common Crawl at scale: streams WARC, WET, and WAT archives from a paths manifest, hands every record to a single processing hook, and keeps disk usage bounded while it runs.

    github.com

  • Photography Portfolio

    Fine art photography portfolio showcasing creative visual work and artistic expression.

    fineartamerica.com