evidently-ai

A complete end-to-end machine learning pipeline to predict employee salaries (USD) using job-related features. Includes automated EDA, robust preprocessing, ML model training with MLflow, drift detection with Evidently AI, and a Flask-based web app for both single and batch predictions.

Professional Data Pipeline is a Python-based data quality and monitoring project that ingests DVC-backed datasets, validates tabular business rules with Great Expectations, and tracks drift in MLflow using Evidently.

Enterprise-grade AI Governance Control Tower on Azure Databricks. Features active zero-trust security (Unity Catalog RLS/masking), data quality circuit breakers (Great Expectations), model observability & drift detection (Evidently AI), and immutable Delta Lake audit logging via GitOps (DABs).

End-to-end MLOps system for probabilistic electricity demand forecasting across 5 Moroccan cities — LightGBM · Airflow · MLflow · FastAPI · Docker

End-to-end MLOps pipeline for Fraud Detection featuring automated CI/CD/CT, MLflow tracking, Evidently AI drift monitoring, and serverless AWS ECS Fargate deployment via CloudFormation.

Industrialisation du modèle de scoring crédit, intégrée dans le projet global de plateforme de scoring MLOps.

End-to-end MLOps pipeline for news classification — experiment tracking with MLflow, data versioning with DVC, FastAPI serving, drift monitoring with Evidently AI, and a 4-job GitHub Actions CI/CD that builds and pushes to DockerHub on every commit.

A secure, scalable FinTech platform with FraudSense (SHA-256 model signing), RegGuard (pgvector/ChromaDB compliance RAG with rate limiting), and FinLens (sanitized SQL agent). Deployed on a 3-replica HA Kubernetes cluster, utilizing Apache Kafka for asynchronous scoring to reduce fraud scoring false positives to 4.7%.

La plateforme de scoring automatisé avec MLOps vise à fournir un système intelligent, scalable et traçable pour : Évaluer le risque de crédit des prospects, Détecter anomalies et fraudes dans les transactions, Automatiser les rapports de conformité (KYC, AML).

⚙️ End-to-End MLOps Pipeline for Credit Risk — FastAPI model serving · Docker containerisation · MLflow experiment tracking (Gini >0.40 promotion gate) · Evidently AI drift monitoring · GitHub Actions CI/CD · GCP Cloud Run · Airflow DAG. Finance-domain MLOps.

End-to-end fraud detection ML pipeline with drift detection, SHAP explainability, and CI/CD

End-to-end MLOps pipeline for loan default prediction with automated retraining, data drift detection, model monitoring, MLflow tracking, Airflow orchestration, FastAPI deployment, and CI/CD automation.

End-to-end credit risk ML system — RF ROC-AUC 0.774 | SHAP explainability | MLflow | Evidently AI | FastAPI

Credit risk platform using an XGBoost + LSTM ensemble with ONNX Runtime INT8 quantization, orchestrated by a LangGraph multi-agent workflow (retrieval, classification, escalation) with Claude structured outputs for explainable credit decisions, deployed on AWS EKS/SageMaker with Feast feature serving and automated drift-triggered retraining.

End-to-end fraud detection system with MLflow, FastAPI, drift detection, and automated retraining

Production-grade ML monitoring system for credit card fraud detection — auto-detects data drift, concept drift & performance degradation, then retrains the model without human intervention.

ML monitoring service with data drift, model performance metrics, alerts, dashboard, PostgreSQL, Docker, and CI/CD.

Scalable MLOps platform for real-time drift detection and monitoring across multiple ML models. Built with a microservices architecture using FastAPI, Qdrant (Vector DB), Evidently AI, and Prometheus/Grafana

Self-healing credit risk pipeline with automated XGBoost tuning via W&B, defensive FastAPI serving, and Dockerized monitoring for closed-loop retraining via Evidently AI.