Analytics & BI Leader · Applied AI
Better decisions, built to last.
I focus on business problems and the commercial outcomes that solving them should deliver. I architected and delivered a regional BI and automation platform that replaced weeks of manual consolidation, and put an inspection app in the hands of more than 100 users across 12 countries. I work with senior leaders to agree what matters, then build what people rely on.
Independent work, built with AI agents under my direction
BUSINESS INTELLIGENCE
Renewable power across the 22 Arab League countries, in English and Arabic. Every Arabic page is generated from its English twin, and the semantic model is what makes the two languages work.
APPLIED AI
Help deciding on a record while standing in a shop: do I own it, what style is it, and what on my shelf is closest. A language model picks matches only from my own collection.
MEDIA PIPELINE
Vertical Instagram reels from phone clips, photos and record audio. I approve a storyboard of real rendered frames before the final video is rendered.
make worker NAME=fix-search TIER=medium
make status
make finish NAME=fix-searchAGENT TOOLING
The working method for running several coding agents at once: one lead, workers in their own worktrees, and review by a second model provider.
01 / 04
Business understanding. Technical delivery.
My career spans commercial finance, analytics consulting and regional BI leadership. I connect the business question to the data, tools and people needed to answer it.
I design how source systems feed trusted, reusable data for reporting, with separate development, test and production environments.
At JLL I architected and built a regional reporting platform in Microsoft Fabric. Reusable dataflows and shared fact and dimension tables replaced weeks of manual consolidation and kept common business definitions consistent across reports.
I build Power BI models people can trust, and reports that answer the question they were built for.
At Keyrus I rebuilt an inherited Power BI model whose relationships produced ambiguous results. I simplified the dimensions and introduced bridge tables, removing the problematic bidirectional relationships and improving performance.
I replace spreadsheets and email chains with apps and flows that keep records and follow up on their own.
At JLL I took a mobile inspection prototype from Excel to Dataverse and expanded it for 65 sites in 12 countries, in four languages. Automated reminders and weekly reporting helped managers see missed inspections and target follow-up.
I direct AI coding agents and check their work with requirements, tests, evaluation data and review by a second model provider.
At JLL I delivered English/Arabic Power BI reporting to production using AI coding agents, reusable skills and human validation. An Arabic-speaking colleague checked the translations; I directed the development and reviewed the outputs.
I work with senior leaders to agree what to build, in what order, and what success looks like.
At JLL I turned workstream needs into an annual BI and automation roadmap, agreed with the client before delivery. I reviewed progress and dependencies monthly, recommending changes to scope or timing when priorities shifted.
I lead delivery teams and help people grow.
At Deloitte I led two data engineers and two analysts delivering marketing analytics, turning stakeholder requirements into team responsibilities and tracking delivery through Scrum. At Deloitte and Keyrus I also managed consultants’ development through coaching, feedback and performance reviews.
I connect analysis to price, margin, cost and return.
At Brown-Forman I assessed pricing and campaigns with Sales and Marketing, weighing margin and break-even against brand objectives. I made recommendations with clear P&L implications, including when a campaign could be worth running without breaking even.
I build things other people can run after I leave.
At JLL I documented data lineage, dependencies and maintenance procedures so others could support the solutions I built. I also wrote Power BI guidelines and a new-starter pack, and led two-day Alteryx and Tableau workshops at Keyrus.
Four personal projects built with AI coding agents under my direction. I set the requirements, made the decisions and tested the results; the agents wrote the code.
Regional BI and automation lead for JLL's Microsoft GWS account across Southeast Asia, Japan, Korea, Australia and New Zealand. I personally delivered the account's reporting and apps, and chaired the global BI workstream for about three years.
A short part-time contract, delivered outside my full-time JLL role, building FP&A reporting to the FP&A manager's scope.
Led a four-person engagement team of data engineers and analysts on digital marketing analytics for a telecommunications client, and managed two consultants' development.
Led data visualisation teams across financial services, FMCG and government, managed two direct reports, and was lead trainer for advanced Alteryx and Tableau.
Data-quality and advanced analytics for a telecommunications infrastructure organisation.
Four roles on a portfolio of about 12 brands, including Jack Daniel's: Inventory Analyst, Commercial Finance Analyst, Commercial Finance Manager (six-month secondment) and Senior Insights Analyst.
Managed about $40 million of pharmaceutical inventory with planning teams and third-party logistics providers.
Education: MCom Business Analytics, University of Sydney (2017); BSc Pharmacology, UNSW (2010).
My agentic set-up · October 2026
My current set-up: multi-provider, modular and transferable. It will evolve over time.
Omarchy laptop (Linux): the always-on machine. It runs CrateDigger as a service, reachable only over my private Tailscale network.
Mac (M2 Pro, 10 cores): where I develop, and where reelkit renders on the media engine.
iPhone: CrateDigger in the record shop.
Claude Code, OpenAI's Codex and pi, all reading the same AGENTS.md in each project.
herdr: one terminal tab per agent, and the channel the agents use to talk to each other.
A lead agent on Claude Opus plans, briefs, reviews and merges. Workers are picked by task: Claude Haiku or Sonnet, or Codex luna, sol or astra. A model from the other provider reviews the work.
agent-kit gives every project the same commands, hooks and decision log, and links its skills into all three harnesses. Both laptops stay in step through GitHub.
ship-with-agents, the orchestration skill from agent-kit: the lead plans the tasks, briefs the workers, relays their questions, reviews with another provider and commits.
Data Goblins' power-bi-agentic-development skills for Power BI and Fabric, Claude's dataviz skill, and each project's own skills.
Music is one of my biggest passions: jazz, especially Japanese jazz, alongside classical, electronic, Latin, soul and experimental sounds. Around 70 records, a Rega Planar 3 and a lot of time in record shops.
