AI for Healthcare · Media Consultant

Turning health data into tools people can use

I'm a media consultant with a background in economics and statistics, working towards becoming an expert in AI for healthcare. I combine numbers, clear communication and growing technical skills, and I learn by building things that work.

Python
NumPy
pandas
MySQL
STATA
SPSS
Microsoft Azure
Debugging

ECG

72bpm

SpO₂

98%

RESP

16/min

Simulated signal

Profile

For more than five years I've worked in media consulting for health programmes across East Africa, where the job is to understand a complex subject and make it clear to the people who need it. I'm now applying that same discipline to code and data, with the goal of working where AI meets healthcare.

I also have a background in economics, statistics and accounting through ACCA. Economics taught me there’s no such thing as a free lunch. Statistics taught me to ask for the sample size. With health systems working within limited budgets, I bring both perspectives to AI: What does it cost? Who benefits? And does the evidence hold up?

I'm conversant with Microsoft Azure, I use NumPy and pandas to clean and visualise data, and I design relational databases. I use AI coding tools throughout, and I test and debug what they produce.

Calculators shifted the intellectual effort from performing arithmetic to understanding the problem. I see AI’s potential in much the same way: reducing routine work so we can devote more attention to judgment, creativity, and the decisions that matter.

Skills

Statistics: Statistical modelling, probability and survey design, using STATA and SPSS
Economics: Econometrics, business analytics and socio-economic research
Data: NumPy and pandas for data cleaning and visualisation
Finance: ACCA studies in financial and management accounting, plus budgeting and pricing for my own consultancy
Programming: Python, JavaScript, React and PHP
Databases: MySQL schema design, foreign keys, joins and seed data
Cloud: Conversant with Microsoft Azure
Debugging: Tracing faults to their root cause and confirming each fix with a measurement
AI-assisted development: Building and testing working software with AI coding tools
Web: WordPress block themes, HTML, CSS, accessibility and performance
Communication: Media consulting and explaining technical work in plain language
Research ethics: Informed consent, do-no-harm practice and careful handling of sensitive health data

Selected projects

Health research · Survey design · Data visualisation

Hanging in the Balance

A 2025 rapid assessment by ATHENA Network and UNAIDS ESA on how international donor funding cuts affected adolescent girls and young women in Africa. I co-designed the report, based on a survey of more than 440 young women across 13 countries, and produced its data visualisations for advocacy with UNAIDS and UN Women.

48% had difficulty accessing healthcare
36% reported increased gender-based violence
82% reported mental health impacts

Healthcare data · MySQL 8

Diabetes patient records database

A mock database for hospital clinical staff, covering type 1 and type 2 diabetes care across 16 tables. Clinical records use soft deletes so their history is preserved, and BMI is a calculated column.

staff
patients
diabetes_diagnosis
visits_encounters
vitals
glucose_readings
hba1c_results
medications
prescriptions
insulin_regimen
complications
lab_results
allergies
appointments
care_plan
audit_log

A design exercise. It contains no real patient data and would need a compliance review before any clinical use.

WordPress · PHP · CSS

Rebuilding my website for speed

I replaced a heavy page-builder setup with a custom block theme, using 32 KB of code and no framework, while keeping the design the same. Drag the slider to compare the live homepage before and after.

Measured on the live homepage
Metric	Before	After
Scripts	33	2
Stylesheets	35	2
Elementor references	301	0
jQuery	Loaded	Removed
HTML size	146 KB	105 KB
Active plugins	16	10

See the live site

React · Remotion · Node · Whisper

Automated video captions

For a drinks brand's social campaign, I built graphics and captions as React components, with speech transcribed locally by Whisper. The audio mixed English and Sheng, which Whisper handled badly, so I wrote a script that applies corrections from a separate file and can safely run again.

The script shares each corrected range across its words by length, which keeps the word-by-word highlight in sync. Here is that idea with a sample line.

Corrections live in their own file

A 3-second range, shared out by the letters in each word. Sample text only.

Python · SQL

Python and SQL practice

Small programs that each get one idea working end to end, plus a MySQL property-listing schema connected to Python.

currency_convertor.pyUSD, EUR, GBP and KES from a nested rate table
compound_interest_calculator.pyCompound interest with input validation
email_slicer.pySplits an address, with a validity check
quiz_game.pyMultiple-choice quiz that tracks your score
airbnb_schema.sqlUsers, rooms, foreign keys and joins

Debugging

I find the cause before I change anything, then measure to confirm the fix worked. These are real problems from my own projects.

Unreadable text in dark mode

Problem: headings and links stayed black on a near-black page, with a contrast of just 1.10:1.

Cause: WordPress colour presets compile to fixed values that the dark theme wasn't overriding.

Result: no contrast failures, with a worst case of 5.93:1.
Oversized images

Problem: 1440px images were being served into 460px slots.

Cause: the page builder emitted sizes="100vw", telling browsers the images filled the screen.

Result: the homepage image payload fell by roughly 90%.
A sticky header that wouldn't stick

Problem: the header scrolled away with the page.

Cause: its parent was exactly its own height, leaving it no room to travel.

Result: moving the rule to the outer element fixed it.

Contact

I'm keen to connect with people working on AI in healthcare.