Roger ThomasDocument Intelligence Architect

Roger Thomas · Ireland

Document Intelligence Architect and Python engineer.

I am a software engineer with a background in astrophysics and scientific computing, specialising in Python backend and data-intensive systems. At Nitro, I design and build document-intelligence services spanning machine learning, generative AI, document indexing and automation workflows.

View projectsView résumé

Background

More than a decade building software with Python.

My experience spans document intelligence, retail analytics, scientific computing, distributed data processing, and SaaS platforms. I have worked as an architect, technical lead, and software engineer, with a focus on backend systems and technical delivery.

Current role
Principal Engineer (Document Intelligence Architect), Nitro
Core areas
Python, backend systems, document intelligence, retrieval/RAG, AWS and Azure
Education
M.Sc. High Performance Computing

Selected projects

Developer tools and full-stack products built to solve concrete problems.

Project 1

In development

jero

A high-performance, msgspec-first ASGI framework for modern Python.

Class-based routing, typed request binding, strict startup validation, and explicit application wiring—without decorator-heavy APIs or a traditional DI container.

  • Python
  • ASGI
  • msgspec
  • Granian
OverviewHomepageGitHub

Project 2

In development

yeetr

A tiny, typed, signature-driven CLI runner.

Run a normal Python function as a command-line tool. Function signatures define arguments and options, with rich help, async support, structured types, and executable scripts.

  • Python
  • Type hints
  • Async
  • CLI
OverviewHomepageGitHub

Project 3

In development

lothc

A typed HTTP client for Python, built on Rust-backed pyreqwest.

Familiar, ergonomic HTTP calls with first-class typed request and response support for pydantic and msgspec.

  • Python
  • pyreqwest
  • Rust
  • pydantic
  • msgspec
OverviewHomepageGitHub

Project 4

Live

two4u

A mobile-only sports prediction game for friends.

A full-stack application that replaces the paper-and-cash score predictions my friends and I make before Ireland international rugby matches.

  • TypeScript
  • SvelteKit
  • Tailwind CSS
  • Skeleton
  • Python
  • FastAPI
  • AWS
  • DynamoDB
  • Terraform
  • Cloudflare
OverviewLive app

Project 5

Live

TaglyDB

A schema-less JSON and blob store, queried by tags.

A schema-less JSON and blob storage API—tag data on write, query by tag combinations instead of designing tables upfront.

  • Cloudflare Workers
  • Hono
  • D1
  • Next.js
  • TypeScript
  • Better Auth
OverviewLive app

Experience

Nitro, Oliver Wyman, and Creme Global.

Roles across document intelligence, retail SaaS, analytics architecture, distributed processing, and scientific risk modelling—from software developer through technical lead and architect.

Professional background

Résumé

Experience, education, and technical background.

View résumé