Operations Analyst, Pretoria, South Africa

Operational data,
read properly.

I work where operations and data meet, turning trip logs, claims records, stock movements and throughput figures into decisions someone can act on. SQL and Python for the analysis, dashboards for the people who have to use it.

See the work LinkedIn GitHub

85,410fleet trips analysed

100,000claims records modelled

12+end to end projects

18 mohands on analytics

What I bring to an operations team

Three things, in the order they usually matter on the job.

Getting the data right

Extracting from relational databases, joining across tables, and cleaning records that were never entered cleanly in the first place.

MySQL, SQLite, window functions, Python, Pandas, Power Query

Reporting people use

Dashboards and reports built around the decision being made, not around the dataset. Utilisation, cost per route, throughput, downtime, exceptions.

Power BI, Tableau, Excel, Streamlit, Matplotlib

Spotting the risk early

Classification and trend models that flag where cost, delay or incident risk is building up, with the drivers explained in plain terms.

Scikit-learn, Random Forest, feature importance

Selected work

Two projects that show the full run: raw operational records in, a decision out.

85,410trips, $298.6M revenue

Fleet and logistics

Logistics fleet operations analysis

Built an analytics pipeline over a 14-table SQLite database covering three years of trips, drivers, maintenance and safety records.
Wrote 15 structured SQL queries on route profitability, driver risk profiling and fleet compliance.
Isolated the highest-risk driver at 7 incidents and $117K in claims, and identified DOT violations as the most frequent incident type.
Philadelphia to Seattle came out as the top route at $10.07M; preventive maintenance was the largest cost category at $963K.

SQL, Python, Matplotlib, Streamlit, repository

99.58%classifier accuracy

Risk and cost

Healthcare claims risk modelling

Cleaned and queried 100,000 claims records in MySQL, then engineered 51 features across demographic, clinical, lifestyle and claims data.
Trained a Random Forest classifier to flag high-cost patients, reaching 0.9999 ROC-AUC, with age, chronic condition count and smoking status as the leading drivers.
Delivered three Tableau dashboards: executive KPIs, risk patterns by segment, and claims cost drivers.
Current smokers averaged 56% higher cost than non-smokers; over-65s cost 62% more than under-18s.

MySQL, Python, Scikit-learn, Tableau, repository

Project index

The rest of the portfolio, by domain.

Project	Domain	What it answered	Stack
Global layoffs analysis	Workforce	Which industries and countries absorbed the deepest cuts, and when the waves peaked	MySQL, window functions
Global AI job market and salary trends	Workforce	How pay and demand vary by role, region and seniority	Python, Power BI
Top 100 SaaS companies, 2025	Finance	Whether funding raised actually tracks revenue and workforce efficiency	SQL, Excel
Company financials dashboard	Finance	Quarterly revenue, cost and margin movement by business unit	Excel, Tableau
Cybersecurity attack analysis	Risk	Which attack types and targets recur across 13,867 logged incidents	MySQL, Python, Streamlit
Restaurant orders analysis	Retail operations	Peak service periods and which menu items carry the revenue	MySQL
Coffee sales analysis	Retail operations	Seasonal revenue patterns and product-level performance	Excel, Tableau
Bike sales, buyer behaviour	Retail operations	Which demographic and income segments actually convert	Excel
Rhythm City Records	Case study	Where genre revenue imbalance came from and how to rebalance stock	SQL

Background

I am Sibusiso Deven Mbuyane. I came into analytics from the workshop floor rather than from a lecture hall, and it shapes how I work: I know what a shift log, a load sheet and a maintenance record look like before they become rows in a table.

Over roughly 18 months I have built 12 or more end to end projects on real and public operational data, each one taken from extraction through to a dashboard or model that answers an actual question. Alongside that I support a family-run bakery operating across Richards Bay and Empangeni, covering stock, merchandising and record keeping, which is where the habit of checking a figure against reality came from.

I hold an N3 in Fitting and Machining, PSIRA grades E, D and C, and a Code 10 licence, and I am currently pursuing the Google Advanced Data Analytics Certificate.

Base: Pretoria, Gauteng. Open to Gauteng, KwaZulu-Natal, remote and hybrid
Looking for: Operations Analyst, Operations Coordinator, Fleet Administrator, Data Analyst
Sectors: Logistics and supply chain, mining, fintech
Certified: Google Data Analytics Professional Certificate. Google Advanced Data Analytics, in progress
Also holds: N3 Fitting and Machining, PSIRA E, D and C, Code 10 licence

Get in touch

Open to permanent roles, contract work and internships. If you have operational data that nobody has had time to look at properly, that is the conversation I want.

Emaildevenmbuyane@gmail.com

Phone+27 73 593 5411

LinkedInlinkedin.com/in/devenmbuyane

GitHubgithub.com/Deven-analyst

Operational data,read properly.