I Built a Real-Time Fraud Detection System on AWS (Full DevOps + MLOps Pipeline)

I built PaySense — a production-grade, real-time ML fraud detection system deployed on AWS
using a fully automated DevOps + MLOps pipeline. In this video I walk through the entire system
end to end: from writing the ML model, to streaming transactions through Kafka, to deploying
on Kubernetes with GitOps, to watching it all on a live Grafana dashboard.
This is the kind of project that belongs on your resume and holds up in any interview.
🔗 GitHub Repo: https://github.com/Yash-Rathod/PaySense2

📌 WHAT YOU'LL SEE IN THIS PROJECT
✅ ML model trained with XGBoost + tracked in MLflow
✅ Real-time transaction streaming via Apache Kafka (Strimzi on EKS)
✅ Fraud classifier running 5ms p99 latency in production
✅ DynamoDB storage + FastAPI results API
✅ Full AWS infrastructure provisioned with Terraform (VPC, EKS, ECR, S3, DynamoDB, IRSA)
✅ Helm charts for every service
✅ ArgoCD GitOps — Git is the single source of truth
✅ GitHub Actions CI/CD with AWS OIDC (zero hardcoded credentials)
✅ Prometheus + Grafana live dashboards (TPS, fraud rate, latency, Kafka lag)
✅ Alerting: fraud rate spikes, consumer lag, high p99 — routes to Slack/email
✅ Low cost per session — fully ephemeral infra, terraform destroy when done

🛠️ TECH STACK
• ML: Python, XGBoost, scikit-learn, MLflow
• Streaming: Apache Kafka, Strimzi Operator
• Storage: Amazon DynamoDB, Amazon S3
• API: FastAPI
• IaC: Terraform (VPC, EKS, ECR, S3, DynamoDB, IRSA)
• Orchestration: Amazon EKS (Kubernetes 1.32)
• Package manager: Helm
• GitOps: ArgoCD
• CI/CD: GitHub Actions + AWS OIDC
• Observability: Prometheus, Grafana, AlertManager
• Registry: Amazon ECR
• Containers: Docker, Docker Compose

📚 RESOURCES
• GitHub: https://github.com/Yash-Rathod/PaySense2
• Strimzi Docs: https://strimzi.io
• ArgoCD Docs: https://argo-cd.readthedocs.io
• kube-prometheus-stack: https://github.com/prometheus-community/helm-charts

🔔 If this project helped you — like, comment what you'd add next, and subscribe.

#devops #mlops #kubernetes #terraform #kafka #argocd #grafana #prometheus
#githubactions #awseks #python #machinelearning #cicd #gitops #helm
#devops #mlops #kubernetes #terraform #kafka #argocd #grafana #prometheus
#githubactions #awseks #python #machinelearning #cicd #gitops #helm
#strimzi #mlflow #dynamodb #docker #fastapi #xgboost #aws #cloudcomputing
#k8s #devsecops #infrastractureasCode #iac #sre #platformengineering
#dataengineering #streamingdata #realtimeml #frauddetection #mlpipeline
#awscloud #amazoneks #amazonecr #amazondynamodb #amazons3 #awsdevops
#softwareengineering #backenddevelopment #pythonprogramming #opensource
#devopsengineeer #cloudnative #microservices #containersecurity
#kubernetesoperator #observability #monitoring #alerting #prometheusmonitoring
#grafanadashboard #kafkastreamimg #apachekafka #eventdriven #eventdrivenarchitecture
#mlops2025 #devops2025 #portfolio #portfolioproject #techinterview
#systemdesign #distributedsystems #scalablesystems #productionml
#continuousintegration #continuousdelivery #continuousdeployment
#dockercompose #containerization #gitopsargocd #helmchart
#kubernetestutorial #terraformtutorial #awstutorial #mlpipeline
#frauddetectionsystem #anomalydetection #realtimeanalytics
#datastreaming #bigdata #fintech #paymentprocessing

Видео I Built a Real-Time Fraud Detection System on AWS (Full DevOps + MLOps Pipeline) канала Yash Rathod