Skip to main content
ITUMELENG MODIKOE

Based in Centurion, South Africa

Junior Data
Scientist

Itumeleng Ronald Jr Modikoe

Currently a Junior Data Scientist at Drive24, turning operational, customer and credit-risk data into explainable scorecards, dashboards and decision support — backed by a Business Intelligence degree and SQL, Python and machine-learning practice.

View My WorkDownload CVContact Me
Itumeleng Ronald Jr Modikoe, Junior Data Scientist at Drive24
Scroll

About Me

A Business Intelligence graduate working as a Junior Data Scientist — equally at home in SQL, in a notebook, and in a room full of stakeholders.

Education

Belgium Campus iTversity, Pretoria

Bachelor of Computing majoring in Business Intelligence, NQF Level 8

2023 — 2026

Coursework spans database development, data warehousing, data science, statistics, information systems, software engineering, project management and research methods.

Final-year studentAcademic average: 78.5%Qualification level: NQF 8

How I Think

I am a Junior Data Scientist with a Business Intelligence background and hands-on experience working with large relational datasets across customer applications, contracts, credit risk, lead management and portfolio performance. I build reusable SQL and Python workflows, investigate data-quality issues, develop operational dashboards and evaluate classification models using explainable and statistically grounded methods. I work with risk, operations, sales and development stakeholders to translate complex data into clear metrics, monitoring tools and decision-support solutions.

Also

  • NTT DATA Bursary — Awarded a bursary to study the Bachelor of Computing in Business Intelligence at Belgium Campus iTversity.
  • Dimension Data Saturday School — Selected for the programme on outstanding academic potential in technology.
  • Peer Tutor — Belgium Campus iTversity — supporting students in programming and analytics modules.

How I Work

Grouped by what the work actually involves — no self-rated scores. Everything here is used in professional work, coursework or certified training.

Data Analysis and Statistics

  • Exploratory data analysis
  • Statistical analysis
  • Segmentation
  • Cohort analysis
  • Trend analysis
  • Anomaly detection
  • Hypothesis testing
  • Survival analysis
  • Data cleaning
  • Data validation

Machine Learning

  • Classification
  • Feature engineering
  • Preprocessing
  • Model selection
  • XGBoost
  • CatBoost
  • Logistic regression
  • Random Forest
  • Decision trees
  • Class imbalance
  • ROC-AUC
  • KS statistics
  • Confusion matrices
  • Feature importance

Credit Risk and Decision Science

  • Credit-risk analytics
  • Weight of Evidence
  • Information Value
  • Scorecards
  • Risk bands
  • Ranking rules
  • Affordability analysis
  • Early-risk monitoring
  • Explainable decision support

Programming and Data

  • Python
  • R
  • SQL
  • PostgreSQL
  • T-SQL
  • pandas
  • NumPy
  • scikit-learn
  • Jupyter
  • LLM-assisted workflows
  • SQL Server
  • Oracle
  • Relational modelling
  • Reusable extraction workflows

Business Intelligence

  • Power BI
  • DAX
  • Amazon QuickSight
  • Excel
  • Matplotlib
  • KPI design
  • Dashboard development
  • Data visualisation
  • Funnel analysis
  • Stakeholder reporting

Data Engineering and Platforms

  • ETL
  • Data integration
  • Data transformation
  • Data-quality checks
  • Talend
  • SSIS fundamentals
  • Git
  • GitHub
  • AWS fundamentals

Additional technologies

From coursework and personal builds, not day-to-day data work.

  • HTML
  • CSS
  • JavaScript
  • C#
  • Node.js
  • MongoDB

Career Timeline

Work Experience

  1. December 2025 — PresentCurrent

    Junior Data Scientist

    Drive24 Pty Ltd

    • Extract, clean, join and analyse operational, customer, application, contract and credit-risk data using PostgreSQL, SQL and Python.
    • Build reusable data-extraction and reporting workflows for recurring analysis, dashboards and stakeholder requests.
    • Develop and maintain Power BI, Amazon QuickSight and Excel reporting for approvals, lead conversion, portfolio performance, deliveries, contracts and operational KPIs.
    • Investigate data discrepancies, validate business rules and improve the reliability of analytical datasets used by business teams.
    • Analyse early termination, churn, contract survival, approval performance, lead conversion and customer-risk patterns.
    • Build and evaluate classification models using XGBoost, CatBoost, logistic regression and Random Forest methods.
    • Apply feature engineering, segmentation, Weight of Evidence, Information Value, ROC-AUC, KS statistics, confusion matrices and out-of-sample validation.
    • Translate model outputs into explainable scorecards, risk tiers, ranking simulations and business decision rules.
    • Collaborate with risk, sales, operations and development stakeholders to define KPIs and convert analytical findings into usable reporting and decision support.

Data Science Work

Professional projects from my work at Drive24, written to be portfolio-safe. Source code and internal data cannot be published, so each is presented as a case study: the problem, what I contributed, and what came out of it.

Case Study 01

Credit Risk and Lead Prioritisation

An explainable modelling and scorecard workflow for ranking applications and supporting risk-informed prioritisation.

Problem
Decision-makers needed a consistent way to compare applicants using incomplete, imbalanced and operationally complex data.
Contribution
Prepared analytical datasets, engineered features, tested statistical relationships, developed classification models and translated model evidence into interpretable ranking logic.
Output
Explainable scorecards, risk bands, lead rankings, comparison simulations and stakeholder-ready analysis.
  • Python
  • pandas
  • NumPy
  • scikit-learn
  • XGBoost
  • CatBoost
  • PostgreSQL
  • Excel
Case Study 02

Portfolio Performance and Early Risk Monitoring

Portfolio analytics for understanding termination behaviour, contract survival and emerging customer stress.

Problem
Stakeholders needed comparable performance views across cohorts while accounting for different observation periods.
Contribution
Built cohort-based analysis, aligned observation windows, compared early termination behaviour and investigated changes in customer credit indicators over time.
Output
Monitoring datasets, reusable SQL queries, risk-change analysis and management-ready summaries.
  • PostgreSQL
  • Python
  • Excel
  • Amazon QuickSight
  • Power BI
Case Study 03

Operational and Conversion Dashboards

Business intelligence reporting that tracks lead conversion, application progress, asset assignment, deliveries and contract outcomes.

Problem
Operational teams required consistent KPI definitions and drill-down reporting across fragmented workflow data.
Contribution
Defined metrics, reconciled sources, created reusable analytical datasets and developed dashboard-ready queries and calculated fields.
Output
Power BI, QuickSight and Excel reporting for operational decisions and recurring reviews.
  • PostgreSQL
  • Amazon QuickSight
  • Power BI
  • DAX
  • Excel
Case Study 04

AI Driven Affordability and Decision Support

A data workflow that combines application, pricing, banking and credit indicators to support explainable affordability review.

Problem
Manual review required consistent interpretation of income, commitments and cash-flow behaviour.
Contribution
Integrated multiple data sources, derived affordability indicators and structured explainable decision logic for review support.
Output
Structured analytical features and decision-support outputs designed to improve review consistency.
  • Python
  • SQL
  • pandas
  • Data-validation workflows

Decision support. Not a fully automated approval system.

My Projects

Personal and academic builds where the code is public — an end-to-end credit-risk model, dimensional modelling, normalised database design, and a couple of things from the developer side.

Machine Learning · Credit Risk

Explainable Credit Default Risk

An end-to-end probability-of-default system on 30,000 credit-card clients: engineered repayment behaviour, compared three classifiers on validation PR-AUC, fixed an operating threshold before touching the test set, then ran calibration and fairness diagnostics and shipped a Streamlit scoring app.

ROC-AUC
0.779
PR-AUC
0.561
Precision
67%
Records
30,000
  • Engineered behavioural and affordability features — repayment delay, months late, credit utilisation, payment-to-bill ratio and six-month bill and payment trends
  • Selected on validation PR-AUC rather than accuracy, because defaults are only 22% of the data
  • Held the test set back for a single final evaluation; threshold and feature importance both come from validation
  • Excluded sex from the model and used it only for subgroup diagnostics, with a unit test enforcing the exclusion
  • Python
  • scikit-learn
  • pandas
  • HistGradientBoosting
  • Streamlit
  • pytest
  • GitHub Actions

Portfolio demonstration on a public UCI dataset from Taiwan. Not a production lending system, and not used to approve or decline real applicants.

View Repository (opens in a new tab)
Predicted against actual remaining useful life for 100 held-out NASA test engines, with a perfect-prediction line and a band showing the 14.7-cycle mean absolute error
View Repository for Fleet Predictive Maintenance and Remaining Useful Life (opens in a new tab)

Machine Learning · Prognostics

Fleet Predictive Maintenance and Remaining Useful Life

A remaining-useful-life model for turbofan engines on NASA C-MAPSS run-to-failure data, with an Isolation Forest anomaly score and a Streamlit dashboard that ranks a fleet by maintenance priority.

MAE (cycles)
14.7
RMSE
20.1
R²
0.77
Test engines
100
  • Split by engine rather than by row, so no cycle from a validation or test engine ever appears in training — the failure mode that makes most RUL projects look better than they are
  • Engineered rolling means, rolling standard deviations and short-term sensor trends within each engine, and capped RUL at 125 cycles to stop long healthy periods dominating the fit
  • Compared linear, Random Forest and gradient-boosted regressors on validation RMSE, preferring the compact model when it lands within 1% of the best score
  • Scored separately for abnormal behaviour with an Isolation Forest fitted only on early healthy cycles
  • Python
  • scikit-learn
  • pandas
  • HistGradientBoosting
  • Isolation Forest
  • Streamlit
  • pytest

Portfolio demonstration on simulated NASA aircraft-engine data. Not certified for aviation, vehicle maintenance or any safety-critical decision.

View Repository (opens in a new tab)

Business Intelligence

EcoTrade Sustainability Insights Dashboard

Interactive Power BI reporting for sustainability KPIs across emissions, product performance and customer behaviour, built on a modelled star schema.

  • Designed a star schema with conformed dimensions to keep measures consistent across pages
  • Authored DAX measures for KPI cards and time-intelligence comparisons
  • Implemented row-level security so each stakeholder group sees only its own scope
  • Power BI
  • DAX
  • Star schema
  • Time intelligence
  • Row-level security
View Repository (opens in a new tab)
Entity relationship diagram for the Paws Haven animal shelter database
View Repository for Paws Haven Animal Shelter Database (opens in a new tab)

Database Development

Paws Haven Animal Shelter Database

A normalised SQL Server database for adoptions, medical records and shelter operations, with auditability and role-based access built in.

  • Modelled the schema to third normal form from real shelter workflows
  • Wrote stored procedures, functions and audit triggers for traceable changes
  • Applied role-based permissions and views to keep sensitive records restricted
  • SQL Server
  • T-SQL
  • ERD
  • 3NF
  • Stored procedures
  • Triggers
View Repository (opens in a new tab)

Certificates

Industry and technical training that supports my work across data science, analytics, cloud and business intelligence.

  • Flagship

    IBM Data Science Professional Certificate

    IBM · Coursera

    Python and SQL for analysis, data visualisation, machine-learning fundamentals and communicating insights.

  • IBM Data Analyst Professional Certificate

    IBM · Coursera

    Excel and SQL fundamentals, Python for analytics, dashboard building and reporting practice.

  • AWS Cloud Practitioner Essentials

    Amazon Web Services · Credly

    Cloud concepts, core AWS services, the shared responsibility model, pricing and cost management.

  • Data Analytics Essentials

    Cisco · Coursera

    Data cleaning, descriptive analytics, SQL querying and storytelling with data.

  • Data Science Bootcamp 2025

    Udemy

    End-to-end projects with pandas and scikit-learn, from preparation through to evaluation.

  • Introduction to Data Science

    Credly

    Data wrangling, visualisation and presenting actionable insights.

  • Mastering SSIS 2016 — Part 1

    Udemy

    ETL pipeline design, packages, data flows and SSIS best practices.

  • Excel 202

    Excel Academy · SABPP CPD

    Intermediate Excel, completed May–July 2026. 2 CPD points, aligned to the SA Board for People Practices competency model.

  • Excel 101

    Excel Academy · SABPP CPD

    Foundational Excel, completed May–June 2026. 2 CPD points, aligned to the SA Board for People Practices competency model.

Get In Touch

I am open to Junior Data Scientist, Data Analyst, BI Analyst, BI Developer and Graduate Data Engineer opportunities.

Contact Info

Find me online

Download CV

Let us work with data that matters

I am open to the roles below. The fastest way to reach me is email or LinkedIn.

  • Junior Data Scientist
  • Data Analyst
  • BI Analyst
  • BI Developer
  • Graduate Data Engineer