Data Engineer at Rubyx, working on lending analytics for financial institutions. On this profile you'll find what I build around one theme I care about: data & AI for financial inclusion in Africa 🌍
Every repo below is a real, working stage of the same story — from raw data to a decision in production:
flowchart LR
A["📦 Raw data<br/>World Bank API · CSVs<br/>transactions · images"] --> B["🌀 Orchestrate<br/>Apache Airflow"]
B --> C["🧱 Transform & test<br/>dbt · DuckDB"]
C --> D["🤖 Model<br/>LightGBM · PyTorch · SHAP"]
D --> E["⚡ Serve<br/>FastAPI · Docker"]
E --> F["📱 Product<br/>APIs · Flutter apps"]
| Project | The story | Stack |
|---|---|---|
| ⚡ BNPL Risk Engine | Real-time Buy-Now-Pay-Later risk scoring API — rule knockouts + LightGBM decisioning (ROC-AUC 0.85), fully tested & Dockerized | FastAPI · LightGBM · Docker |
| 🌍 West Africa Financial Inclusion | Real-data ELT pipeline on financial inclusion across WAEMU countries: World Bank API → dbt → BigQuery → dashboard | Airflow · dbt · BigQuery |
| 🌾 GblackAI | AI crop pest & disease detection with Wolof voice support, built for AbiHack 2025 — live API docs ↗ | PyTorch · FastAPI · Flutter |
Pipelines that run on schedule, recover from failure, and don't wake anyone up at 3am.
| Repo | What it demonstrates |
|---|---|
| airflow-dynamic-dag-factory | YAML-driven DAG factory — config-as-code pipelines instead of DAG proliferation |
| airflow-batch-credit-scoring | Daily batch credit scoring: chunked dynamic mapping, versioned models, PSI drift gate |
| airflow-dbt-cloud-orchestration | Full orchestration of dbt Cloud jobs — manifest parsing, quality gates, intelligent alerting |
| airflow-cloud-data-platform | Production patterns: dynamic task mapping, Kubernetes, custom XCom backend, cost optimization |
Because a metric nobody trusts is a metric nobody uses.
| Repo | What it demonstrates |
|---|---|
| dbt-advanced-patterns | Incremental models, SCD Type 2, custom macros, slim CI — for fintech data platforms |
| dbt-testing-and-data-quality | Testing framework with custom macros, automated data-quality monitoring, CI quality gates |
| credit-scoring-dbt | Raw fintech transactions → ML-ready credit features, engineered in dbt |
| digital-lending-analytics-dbt | Data warehouse & BI for digital lending — loan portfolio KPIs and cohort analysis |
Applied to the problems fintech actually has.
| Repo | What it demonstrates |
|---|---|
| msme-alternative-credit-scoring | Alternative credit scoring for African MSMEs from mobile-money & behavioral data |
| aml-graph-neural-network | Anti-money-laundering detection with Graph Neural Networks (GraphSAGE) |
| fraud-detection-deep-learning | Credit-card fraud detection: autoencoder + LightGBM, explained with SHAP |
| financial-nlp-news-market | Financial news sentiment → market prediction, with FinBERT & LSTM |
| loan-default-ml-pipeline | End-to-end loan-default prediction pipeline with SHAP explainability |
version: 2
models:
- name: ibrahima_gabar_diop
description: "AI & Data Developer — Thiès, Senegal 🇸🇳"
meta:
current_role: "Data Engineer @ Rubyx"
education:
- "MSc Big Data Analytics · UNCHK (in progress)"
- "BSc Mathematics & Computer Science"
- "Sonatel Academy alumnus"
competitions: [Zindi, Kaggle, GalsenAI, Devpost]
happy_to_talk_about: [data engineering, LLMs, computer vision]
columns:
- name: curiosity
description: "How things work under the hood"
tests:
- not_nullData & Analytics
AI / ML
Backend & Infra
Frontend & Mobile
