Data & Automation Engineer
I engineer scalable data infrastructure, pipelines, dashboards and automations that leadership teams run on and companies grow from. Five years across health distribution, IT services and international development.
Abuja, Nigeria · WAT (UTC+1) · Open to remote work and relocation
Personal AI Job Board
Oze probably found your job posting before you found me.
Oze is an AI job recommendation engine that searches five job markets worldwide and scores every opening against my CV.
Good roles were buried among thousands of unsuitable listings, and checking job sites by hand took hours I didn't have. So I built a self-running AI lead-generation pipeline: about 1,500 listings a week, narrowed to the 15 worth my time, waiting on my phone before I wake.
Read the case study →Automated ERP Reporting
An end-to-end pipeline that catches sales and stock reports straight from email, cleans them, and feeds a single live Power BI app built on Microsoft Fabric.
Reports arrived from several locations and two different ERP systems, each with its own layout: buried headers, merged cells, mismatched column names. Stitching them together by hand took hours every week and let errors reach the numbers leadership decided on. Now every file is captured, parsed and checked automatically, the app refreshes without anyone touching a spreadsheet, and leadership reviews numbers they can trust.
Read the case study →Inventory Intelligence System
An inventory and analytics system built around Plantanova, a Spanish exporter of capers, olives and pickles, entirely on Google tools.
Perishable stock, cold rooms, and a different shelf life for every product: one missed expiry or empty shelf can cost an export order. I modelled the whole operation, tracing every batch from supplier to customer, shipping oldest-expiry-first, re-ordering before stock runs short and alerting the right manager before trouble lands. On two years of modelled trading it holds 97 to 100% order fill with cold rooms around 60% full.
ISO Compliance on Microsoft 365
A Microsoft 365 system replacing manual quality, HSE and document control for an offshore EPCI contractor.
Risks, incidents, corrective actions and permits sat in separate registers across ten departments, so proving ISO and API compliance meant chasing emails and spreadsheets. I built the apps, workflows and reporting on top: every record numbered, scored and logged automatically, a high risk left open 14 days raises its own corrective action, and one Power BI suite, secured by role, shows leadership where the company stands.
Read the case study →Salon Retention Analytics
A churn and retention analytics system for Aura Beauty Salon in New Zealand, from database to live Power BI app.
Salon clients rarely cancel; they just stop booking, and the owner finds out when revenue dips months later. I built the stack end to end off the Zenoti API: salon records funnelled into a cloud MySQL database, SQL views that classify every client as active, at risk or churned each month, and a Power BI app that shows who is slipping, which services keep clients, and who can still be won back.
Year at a Glance
A Google Sheets to Google Calendar automation that lets one person run another's diary without ever entering their calendar.
An executive's calendar is private, but the assistant who runs it needs somewhere to plan, move and clear commitments. YAAG gives them a year-at-a-glance sheet: fill in a day and the event appears, change it and the event moves, clear it and the event is gone. The executive only ever opens the calendar.
Read the case study →I build things that work, on screen and off it.
On screen, that means data systems people actually run on: pipelines that catch reports before anyone downloads them, dashboards that show leaders where the business stands and what needs attention next, and automations that take repetitive work off people's desks for good. Five years across health distribution, IT services and international development have taught me the same lesson each time: a number only matters if someone trusts it enough to act on it. That is why I automate first: hours and money leak fastest through manual work, and people forget, but systems remember.
Off screen, I'm a hands-on DIYer: electrics and electronics, plumbing, masonry, woodwork and carpentry, metalwork, arts and crafts. Mostly home improvement and interior decoration, and mostly because I love finding out how things work.
I hold an MSc in Geographic Information Systems and Udacity certifications as a Data Scientist and Data Analyst. I also teach, from industry leaders to young data enthusiasts, showing them what is possible with data.
Case study · Business Intelligence · 2026
Automated ERP reporting. It catches sales and stock reports the moment they arrive by email, reconciles two very different systems into one clean model, and keeps a live Power BI app current without anyone touching a spreadsheet.
The main path in orange runs unattended. Anything that fails a check is set aside and reported, never silently dropped.
Sales and stock reports arrived by email from several locations, exported from two ERP systems that agreed on almost nothing. One produced clean tables. The other produced report-style sheets, with metadata blocks on top, merged cells, headers that moved from file to file, and its own names for the same columns and products.
Every cycle meant downloading attachments, cleaning each format by hand and stitching the figures together in spreadsheets. It took hours, left no record of where a number came from, and let errors travel all the way into the report leadership used to decide.
Speed was never the real goal. A faster report nobody trusts changes nothing, so every record carries the file it came from and the run that processed it, and anything that fails a check is quarantined and reported instead of quietly dropped.
Every number in the app can be traced back to the file it came from.
Case study · Business Intelligence · Google · 2025
Inventory intelligence for a Spanish food exporter. Google Sheets holds the data, Apps Script runs the logic, and Looker Studio shows the business where it stands, with no paid software anywhere in the stack.
Everything in orange runs on its own schedule. Nobody has to open the spreadsheet for the dashboard or the alerts to stay current.
Plantanova makes and exports capers, olives and pickled vegetables. In the operation I modelled around it, stock sits in a main warehouse and two cold rooms, every product has its own shelf life and temperature range, and export buyers reject anything too close to its expiry date.
Run on spreadsheets and memory, an operation like this goes wrong in predictable ways. Old lots get forgotten until they expire, reorders happen when someone notices an empty shelf, and margin is only known once the month has closed. I set out to design the system that would stop each of those before it happens.
The dashboard is live and fully interactive. Open it full screen ↗
A dashboard is only as good as its arithmetic. Cost of sales comes from what each shipped batch actually cost, not from what was bought that month. Stock is valued at cost, space is measured in the same units as the rooms, and every alert is written to a log before any email goes out. When I audited the build, those were exactly the traps I found and closed.
Every kilo shipped traces back to its batch, its supplier and the room it left from.
Case study · Low-code · Microsoft · 2026
An ISO compliance system on Microsoft 365 for an offshore EPCI contractor. Power Apps captures the record, Power Automate runs the process, and Power BI shows leadership where the whole company stands.
Records move through the orange path on their own. Nobody chases an approval, numbers a record or compiles a report by hand.
The contractor was working toward ISO 9001, 14001 and 45001 and API Q1 and Q2. Its risks, incidents, corrective actions, permits and documents sat in separate registers across ten departmental sites, and approvals were chased by hand. Proving compliance to an auditor meant rebuilding the story from emails and spreadsheets.
A colleague had provisioned the SharePoint foundation. I built what makes it run: the entry apps, the workflows, the audit trail and the executive reporting. What is demonstrated here is the pre-production build, before the system went live.
Exported from Power BI, with the client's name and staff names removed. Open the PDF ↗
In a compliance system, the record is the product. Every workflow writes to the same audit logger, the lists the workflows depend on are created by PowerShell scripts that can run twice without breaking anything, and each Power BI view is mapped to the Azure AD group allowed to see it.
Every action leaves an audit record, written by one logger every workflow shares.
Case study · Automation · Google · 2025
Year at a Glance. A planning sheet that runs a private calendar: the person who plans the diary works in Google Sheets, and the person who lives by it only ever opens Google Calendar.
The assistant works on the left, the executive lives on the right, and the calendar in between is the only thing they share.
A senior executive's calendar holds everything: board meetings, travel, family, private appointments. The person who actually runs the diary, an assistant or chief of staff, needs somewhere to plan, move and clear commitments. Giving them full access to the calendar exposes all of it. Running it through messages means double entry, missed changes and events that should have been cancelled weeks ago.
YAAG started as my own year plan. Every project on this page began as a line on one of its dates.
YAAG writes to one dedicated calendar and nothing else. The executive shares that calendar with whoever runs the sheet and lays it over their own, so the assistant can manage every commitment without ever seeing the main calendar or anything else in it.
The assistant manages the plan. The calendar, and everything else in it, stays the executive's.
Case study · Business Intelligence · SQL · 2025
Churn and retention analytics for Aura Beauty Salon in New Zealand. Built off the Zenoti API: salon records funnelled into a cloud MySQL database, SQL views that do the heavy lifting, and a Power BI app that shows which clients are drifting away before they are gone.
Most of the logic lives in SQL, not in the dashboard, so every figure can be queried and checked on its own.
A salon lives on repeat visits. When a regular drifts away there is no cancellation and no complaint, only a gap in the diary that nobody notices until revenue dips months later. By then the client has found someone else.
Aura needed to see that drift while it could still be reversed: who is slipping, which services and staff keep clients coming back, and whether lapsed clients can be won back. The data architecture is built off the Zenoti API, the salon software: records are funnelled from the API into SQL, so the same pipeline plugs into any Zenoti account. The figures shown are modelled.
Live and fully interactive. Open it full screen ↗
The dashboard is the last step, not the engine. Relationships and constraints live in the cloud MySQL database, the heavy aggregation lives in SQL views, and Power BI only reads results that are already correct. Change a rule once in SQL and every page follows.
Every number in the report comes from a SQL view that can be run and checked on its own.
Case study · Automation · 2026
Personal AI Job Board. It reads every new opening so I don't have to, scores each one against my CV, and sends me only the few worth my time.
Anyone who has searched for a job knows the grind: the same roles posted across many sites, different filters on each, and the few genuine fits buried under thousands that aren't. Reading them properly takes hours every week, and the best opportunities are still easy to miss.
More searching isn't the answer. What's needed is something that reads everything, judges each role the way a careful person would, and only interrupts when a job is worth the time. That is what Oze does.
Not a single scoring failure across all 1,634 jobs. Nothing below the cap is lost either: every scored job stays in the log.
A job source that quietly returns nothing looks exactly like a slow market, so Oze checks itself on every run, raises an alarm when anything breaks, and stops outright when data arrives in a shape it doesn't expect.
When Oze is silent, the market is quiet. It never means something broke.