Kofi E. Ampaabeng

Selected Work

20 systems I have architected, built and run: enterprise platforms, consumer apps, and the tools and infrastructure behind them. 20+ years across regulated banking, data platforms and applied ML.

20
systems
6
enterprise
7
ML · LLM
20+
years

07 · Regulatory data

RegAuth

Quantification is the point.

A scrape → parse → NLP → index → serve pipeline that turns the U.S. federal legal corpus into structured, searchable data.

  • Enterprise

Surfaces

  • Web
  • Self-hosted
  • API
  1. 01Seven corpora, one pipeline — CFR, U.S. Code, Public Laws, Executive Orders, Statutes, the Federal Register and Acts.
  2. 02Distributed processing — PySpark exports NLP features to partitioned Parquet with tuned compression, pushdown and partition sizes.
  3. 03Scheduled workers — Celery workers on two queues, one for browser-driven scrapes, with a nightly re-index.
  4. 04Search and serving — a Postgres → Solr indexer behind an Express BFF and a React SPA.
  5. 05Observability — opt-in Sentry error and performance reporting.
  • 2 Celery queues
  • Spark to Parquet
  • Solr search
SurfacesWebSelf-hostedAPI
SOURCESCOLLECTPROCESSNLPSTORESERVEindexStatutes and codesUSC · STAT · PLAW · ACTSRegulationsCFR · EO · FRBeat scheduleScrape workersdefault + browser queuesObject storagelocal or S3Parserspandas · SparkProcessorsauthority links · featuresSpark exportpartitioned ParquetSummariesBART · LEDPostgreSQLSolrExpress BFFReact SPA
Architecture, drawn from the codeServiceStoreModel usedModel trained hereFeedback
  • PySpark
  • Parquet
  • Celery · Redis
  • PostgreSQL
  • Apache Solr
  • BART
  • React · Express