Open to new opportunities · UK-based

Data engineering · Machine learning · Automation

I design and build the data platforms businesses run on.

End-to-end ownership: I scope the problem, design the data model, write the code, integrate the systems, build the reporting, and run it in production. Twenty-five years turning fragmented, manual processes into reliable, intelligent data platforms.

See my work Get in touch
0
Years in data & analytics
0
Company record-pairs matched & scored
0
Customer records handled at scale
0
Faster: a critical ETL, 8h → 2 min
Selected work

What I've built

Both flagship systems below were originated unprompted, designed end-to-end, and run in production today. Names are withheld while I'm still at my current employer; I'm happy to walk through either in detail.

Global commission-management platform

SQL ServerAdvanced T-SQLPower BIAutomation

A global platform that runs sales-commission calculations across multiple tiers for an entire technology and services sales organisation. I designed the SQL Server data model, wrote the complex T-SQL behind the calculations, built the reporting layer senior stakeholders use, and automated the data ingestion. It grew from a small reporting request into the platform it is now, through a series of calls I made as the real need became clear.

Relied on byFinance + Sales Ops, daily
Calculation errorsClose to zero
OwnershipMine, end-to-end

Company-data mastering pipeline

PythonLightGBMSentence embeddingsLLM (grounded)SQL Server

An entity-resolution pipeline that gives every company account across the business a single canonical identity, so that every spelling, abbreviation, subsidiary, and brand of the same parent rolls up together. A multi-stage Python pipeline doing exact and fuzzy matching, machine-learning classification with a model I trained, and LLM verification using grounded search, all sitting on SQL Server. Around a million record-pairs are matched, scored, and routed through it. I taught myself the ML and LLM engineering to build it.

Pairs matched~1 million
PipelineSix stages + consolidation
ML & LLMSelf-taught, in production

Automated BI security framework

SQL ServerRow-level securityBI

An automated row-level security system linking SQL Server and the BI layer, giving secure, auditable, scalable access control across reporting environments without manual upkeep.

Self-service marketing analytics platform

Dimensional modellingAccess control12M+ records

A self-service analytics model letting non-technical marketing teams safely segment over 12 million customer records, scaling one campaign from 12 to 200+ tailored variants with full data integrity throughout.

Toolkit

The stack I actually use

All of it in production in something I've built or maintain. The machine-learning and LLM pieces are self-taught from scratch.

// DATA & SQL

Data & SQL Server

  • Advanced T-SQL
  • Stored procedures
  • Query optimisation
  • Execution-plan tuning
  • Dimensional modelling
  • Star / snowflake schemas
// PYTHON & ML

Python & machine learning

  • Python
  • pandas
  • LightGBM
  • Sentence transformers
  • Embeddings
  • Model training
// AI & LLM

LLM integration

  • Grounded LLM verification
  • Prompt & pipeline design
  • API orchestration
  • Caching & cost control
// ETL & PIPELINES

ETL & integration

  • SSIS
  • Fivetran
  • REST APIs
  • Custom connectors
  • Batch & incremental loads
// BI & REPORTING

BI & reporting

  • Power BI
  • DAX
  • Sisense
  • Balanced scorecards
  • Executive MI
// PLATFORM & OPS

Platform & DevOps

  • Power Apps (Canvas)
  • Power Automate
  • Git / GitHub
  • Azure DevOps
  • CI/CD
  • PowerShell
How I work

How I operate

The work is autonomous because it has to be. Most days are a steady run of judgement calls, code, and quiet documentation.

01

I originate the work

Both flagship systems started without a brief. I spotted the problem, designed the answer end-to-end, and shipped it. I don't wait to be told what to build next.

02

I own it in production

Scope, data model, code, integration, reporting, and the long tail of running it live. The architecture decisions are mine, and they hold up.

03

I set my own standards

Version control on every stored procedure, change logs, written runbooks, automated tests, and a memory of past bugs so I don't relitigate solved problems. Nobody handed these down; the work needed them.

04

I'm easy to work with

I tailor how I explain things to the audience, take feedback well, and I'm not precious about my work. If someone has a better idea I'll adopt it; if they don't, I'll say so plainly.

"I work long hours, but not because I have to. I don't like leaving a problem half-solved. The upside for you is that I'm reliable on delivery, and the projects I take on tend to ship."

Paul J Brooks
Experience

Where I've done it

Twenty-five years in data, across enterprise technology and regulated financial services, building platforms from the ground up.

Sept 2021 — Present

RWS Group

Senior Data & Innovations Developer

Lead the design and delivery of the data and automation platforms that Finance, Sales Operations, and increasingly the wider business run on. Built production ETL across SQL Server, Python, SSIS, and Fivetran, including rebuilding a critical daily ETL from over eight hours to under two minutes. Delivered a global commission-management platform and a company-data mastering pipeline from scratch, and introduced version control, testing, and documentation standards the team didn't previously have.

2007 — 2021

Lloyds Banking Group

Data, Analytics & MI leadership roles

Fourteen years in a highly regulated financial-services environment. Executive MI Manager for Complaints, producing board-level balanced scorecards and risk dashboards. PPI Analysis Manager, owning the analytics framework for one of the UK's largest financial remediation programmes. Digital Insights Manager, building a self-service marketing-analytics platform over 12 million customer records. Throughout: FCA/FOS regulatory reporting, full audit trails, and zero tolerance for error in board-level MI.

Earlier data roles from 2001 at Harrods, Dun & Bradstreet, and Hallmark Cards, among others. Full employment history available on request.

Education

Brunel University

Product Design BSc (Year 1). A-Levels in Design Technology, Physics, and Computing.

Always learning

Self-taught, in production

Python, ML model engineering, LightGBM, sentence embeddings, and grounded LLM integration, all picked up as the work called for it.

Outside work

Entrepreneurial streak

Ran a wedding-photography business and a woodworking business alongside full-time roles. Comfortable working independently to a deadline.

Contact

Let's talk.

Whether you're hiring, scoping a project, or just want to compare notes, send me a message and I'll get back to you. No detail is too small.

Connect on LinkedIn

Your message reaches me directly. My contact details stay private until I reply.