Skip to content
View muhammed-keita-ml's full-sized avatar
3 followers · 13 following

Block or report muhammed-keita-ml

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
muhammed-keita-ml/README.md

Muhammed Keita — ML Systems Engineer

I build production ML systems with an emphasis on reliability after deployment. Domain focus: FinTech and financial services — where model failure has a measurable cost.

What I ship

# Project Stack Status
07 Cloud-Native Fraud Detection API FastAPI · Docker · AWS ECR · AWS EC2 · GitHub Actions · Kubernetes · MLflow ✅ Live
06 Credit Card Fraud Detection Pipeline XGBoost · scikit-learn · imbalanced-learn · MLflow · DagsHub · Gradio ✅ Live
05 Production ML Monitoring API FastAPI · Docker · GitHub Actions · pytest · pandas ✅ Live
04 Heart Disease Risk Pipeline scikit-learn · MLflow · DagsHub · Hugging Face · GitHub Actions ✅ Live
03 Modular CLI Productivity Tool Python · Modular Architecture ✅ Complete
02 CSV Data Processing Pipeline Python · pandas · NumPy · MLflow ✅ Complete
01 Automated File Organizer Python · os · shutil · json ✅ Complete
08 Financial Risk Intelligence Platform (FRIP) Multi-model · fraud + credit scoring + monitoring 🔄 Planned

Key results

Project Metric Value
Project 07 Live AWS EC2 deployment 108.128.140.230:8000/docs
Project 06 PR-AUC (XGBoost + class weighting) 0.880
Project 06 Recall@P90 0.837
Project 06 Configs with Recall@P90 = 0 despite okay F1 6 of 13
Project 04 AUC (Random Forest) 0.921
Project 04 Recall 0.922

Research orientation

My production work surfaces questions that belong in the literature:

  • When does statistical drift detection fail to predict actual model performance degradation?
  • How do SMOTE variants interact with temporal validation in production fraud detection?
  • Why do standard metrics (F1, accuracy) hide operationally unusable models under extreme class imbalance — and what should replace them?

Current focus: Concept drift × class imbalance × model degradation in real-time fraud detection systems. Project 06 is the first systematic empirical step toward answering this.

Stack

ML/MLOps: XGBoost · scikit-learn · imbalanced-learn · MLflow · DagsHub Serving: FastAPI · Docker · Hugging Face Spaces Cloud: AWS EC2 · AWS ECR · GCP (Professional ML Engineer certified) DevOps: GitHub Actions · pytest · Kubernetes Tracking: MLflow Model Registry · DagsHub Language: Python

Certifications

Certification Issuer
Professional ML Engineer Google Cloud
Machine Learning in Production DeepLearning.AI
Machine Learning Specialization DeepLearning.AI / Andrew Ng
IBM DevOps and Software Engineering Professional IBM
IBM Data Science Professional IBM
Introduction to Model Context Protocol Anthropic
Building with Claude API Anthropic

Now

  • Building Project 08 — Financial Risk Intelligence Platform (FRIP): fraud detection + credit scoring + monitoring as one multi-model platform
  • Open to remote ML Engineer, MLOps Engineer, ML Platform Engineer roles — UK · Netherlands · Germany · Canada · US

Find me

LinkedIn Hugging Face Contra DagsHub Medium

GitHub Stats

Muhammed's GitHub Stats

Pinned Loading

  1. project-07-fraud-detection-k8s project-07-fraud-detection-k8s Public

    Cloud-native ML inference service: FastAPI, multi-stage Docker, AWS ECR/EC2 CI/CD via GitHub Actions, MLflow Model Registry, Kubernetes manifests (EKS-ready), Prometheus instrumentation. 10 pytest …

    Python 2

  2. project-06-fraud-detection-pipeline project-06-fraud-detection-pipeline Public

    13-experiment fraud detection benchmark: 5 imbalance strategies × 3 models. XGBoost + class weighting wins (PR-AUC 0.880, Recall@P90 0.837). Key finding: F1/accuracy hide broken models under extrem…

    Python 1

  3. project-05-heart-disease-api project-05-heart-disease-api Public

    Production ML monitoring system: statistical drift detection, FastAPI serving layer, Docker, GitHub Actions CI/CD, pytest. Research context: monitoring methodology for deployed classification systems.

    Python 1

  4. project-04-heart-disease-pipeline project-04-heart-disease-pipeline Public

    End-to-end ML pipeline with systematic model selection, MLflow experiment tracking on DagsHub, and live Hugging Face deployment. Random Forest AUC=0.921. CI/CD via GitHub Actions.

    Jupyter Notebook 1

  5. 02-csv-data-processor 02-csv-data-processor Public

    Production-style Python CSV processing pipeline. pandas, NumPy, MLflow experiment tracking. Standard-library-first, reproducible, CLI-driven.

    Python 1

  6. 03-cli-productivity-tool 03-cli-productivity-tool Public

    Modular Python CLI demonstrating clean architecture, separation of concerns, and extensible design patterns.

    Python 1