Back to projects
CloudData Science

CloudEco — Wildlife Detection API

Cloud-native ML inference with YOLOv8, FastAPI, Docker, Kubernetes & Terraform

Built and deployed a cloud-native wildlife detection REST API powered by a YOLOv8 model, FastAPI, Docker, Kubernetes, Terraform, and structured load testing.

PythonYOLOv8FastAPIDockerKubernetesTerraformLocustCloud InfrastructureML Deployment
Case Study

Problem

Wildlife monitoring workloads needed a scalable, reproducible way to run object detection inference on cloud infrastructure without manual provisioning.

My role

Cloud / ML Engineer — designed the inference API, container strategy, Kubernetes manifests, Terraform stack, and load-testing harness.

Solution

Wrapped a YOLOv8 model in a FastAPI service, containerised with Docker, deployed to Kubernetes, and provisioned multi-node cloud infrastructure with Terraform. Locust scripts benchmarked latency, throughput, and reliability across configurations.

Challenges

  • ›Balancing inference latency against pod cost across replica counts.
  • ›Reproducible multi-node infrastructure across environments.
  • ›Benchmarking throughput and failure rate under structured load.

Key features

  • ›Deployed a cloud-native ML inference API under constrained cloud resource settings.
  • ›Provisioned multi-node cloud infrastructure using Terraform Infrastructure-as-Code.
  • ›Orchestrated containerised FastAPI services across Kubernetes deployments.
  • ›Conducted Locust load testing to benchmark throughput, response time, failure rate, and scalability.
  • ›Evaluated cloud performance using structured metrics including latency, concurrent traffic, and reliability.

Results

  • ›Documented performance envelope for capacity planning.
  • ›Reproducible Terraform-driven environments.
  • ›Structured load-test evidence for scalability claims.

Cloud-native ML inference

Benchmarked across pod counts

Reproducible IaC deployments