AI/Data Engineer · PySpark · Databricks · Agentic AI · Microsoft Fabric
I build production-grade data pipelines, ML systems, and agentic AI workflows — grounded in a strong foundation in mathematics and statistics.
- Applied MSc in Data Science & AI — DSTI School of Engineering, Nice · First Class / Summa Cum Laude (85%)
- Joint MSc in Mathematical Modelling — MathMods Erasmus · University of L'Aquila · University of Hamburg · Université Côte d'Azur
- BSc Mathematics — Usmanu Danfodiyo University · First Class (4.91 / 5.00)
At Amadeus, I built PySpark ETL/ELT pipelines, a production Transformer-based QSI demand prediction model, and Power BI dashboards on Azure Databricks using a medallion lakehouse architecture. I also automated CI/CD infrastructure on AWS using Jenkins, Kubernetes, and Docker.
Currently focused on agentic AI workflows, Microsoft Fabric, and portfolio-grade data engineering projects.
| Domain | Description |
|---|---|
| Agentic AI | LLM-powered multi-agent systems, tool use, structured output, evaluation-driven orchestration |
| Lakehouse Pipelines | PySpark ELT with medallion architecture, Delta Lake, schema enforcement, incremental refresh |
| ML & MLOps | Scalable model training, experiment tracking (MLflow), DVC pipelines, deployment-ready workflows |
| Analytics & BI | Power BI semantic layers, DAX, Deneb/Charticulator visuals, star schemas |
| DevOps / Cloud | Jenkins, Docker, Kubernetes, Helm, ArgoCD, Terraform on AWS and Azure |
Languages
Data Engineering
-FF694B?style=flat&logo=dbt&logoColor=white){kind=icon}
AI / Machine Learning
Analytics & BI
Cloud & DevOps
| # | Project | Description | Stack |
|---|---|---|---|
| 🤖 | Agentic Project Management Workflow | Multi-agent LLM system transforming product specs into user stories, engineering tasks, evaluation feedback, and execution plans | Python · OpenAI · Agentic AI |
| 🗺️ | Agentsville AI Trip Planner | AI agent system for personalized trip planning — tool use, structured outputs, Pydantic validation, evaluation-driven itinerary revision | Python · OpenAI · Pydantic · Agents |
| 🧪 | MLOps Pipeline with DVC | Reproducible ML pipeline with DVC data versioning, MLflow experiment tracking, and GitHub Actions CI | Python · DVC · MLflow · GitHub Actions |
| 🧠 | Reddit Mental Health NLP Classifier | End-to-end NLP classification pipeline across 16 mental-health categories with a deployable Streamlit demo | Python · Transformers · CatBoost · Docker |
| ☁️ | Azure HDF5 Sync Pipeline | Lightweight data engineering utility — syncs simulation outputs to Azure Blob Storage with incremental logic | Python · Azure Blob Storage · SDK |
| 🚀 | DevOps CI/CD on AWS + Kubernetes | Full CI/CD pipeline: Jenkins on EC2, Docker agents, SonarQube, Helm, ArgoCD / GitOps | AWS · Jenkins · Docker · Kubernetes · Helm |
| 🔊 | wav2vec2 Layer Probing | Research: layer-wise probing of wav2vec2-base to study where spoken digit and speaker information emerge in SSL speech representations | Python · PyTorch · Hugging Face |
| 🔬 | Mouse Lung 3D Segmentation | Research CV pipeline for 3D segmentation of light-sheet microscopy volumes; exported to ITK-SNAP / 3D Slicer | Python · Watershed · K-Means · Image Processing |
| 👁️ | Facial Keypoint Detection | CNN-based facial keypoint detection: Haar cascade face detection + a from-scratch PyTorch model predicting 68 landmarks per face, with a diagnosed BatchNorm bug fix and a live Gradio demo | Python · PyTorch · OpenCV · Gradio |
| 🖼️ | Show and Tell: Image Captioning | PyTorch CNN-RNN encoder-decoder implementing Vinyals et al.'s Show and Tell architecture on COCO, with greedy and beam-search inference, a CLI tool, and full test coverage | Python · PyTorch · CNN-RNN · COCO |
| 🧭 | Graph SLAM: 2D Landmark Tracking | From-scratch Graph SLAM implementation reconstructing a robot's trajectory and landmark positions from noisy motion/sensor data alone, via constraint matrices (Ω, ξ) and least-squares estimation | Python · NumPy · Robotics · SLAM |
| 📉 | Dynamic Risk Assessment System | End-to-end MLOps pipeline for client attrition risk prediction — automated ingestion, retraining, deployment, monitoring, diagnostics, reporting, and drift detection | Python · MLOps · Model Monitoring |
| 💰 | Census Income Classifier (FastAPI + MLOps) | End-to-end MLOps project for Census Income classification with FastAPI, automated + slice-based testing, GitHub Actions CI/CD, and cloud deployment | Python · FastAPI · GitHub Actions · MLOps |
| 🏠 | ML Pipeline for Short-Term Rental Prices | End-to-end MLOps pipeline for Airbnb price prediction using MLflow, Hydra, and Weights & Biases — data validation, experiment tracking, training, evaluation, and reproducible deployment | Python · MLflow · Hydra · W&B |
| 🛡️ | Omnitrainer: Multimodal Moderation | Multimodal AI content moderation for customer-service training — demo notebooks and agents spanning text, image, audio, and video, with a full test suite | Python · Multimodal AI · Agents |
| 🛰️ | NASA Intelligence Chat | OpenAI-powered retrieval-augmented NASA mission chat system with citation-aware answers, ChromaDB retrieval, a Streamlit UI, and RAGAS evaluation | Python · OpenAI · RAG · ChromaDB · Streamlit |
| 🎮 | UdaPlay: AI Research Agent | AI research agent for video game analytics using RAG, ChromaDB, Tavily web search, retrieval evaluation, and conversation memory | Python · RAG · ChromaDB · Tavily |
-0078D4?style=flat&logo=microsoft&logoColor=white){kind=icon}
- Agentic AI workflow systems — multi-agent orchestration with LLM tool use
- Microsoft Fabric / Lakehouse data engineering projects
- PySpark / Databricks production pipeline portfolio
- MLOps pipelines with DVC, MLflow, and CI/CD integration
When I'm not building pipelines or deploying models, I'm mentoring others in tech, playing basketball, or hiking.