Build ML-Driven Java Microservices
Budget: $5,000 – $10,000 USD
The project involves creating brand-new Java-based micro-services that expose and consume machine-learning models. Each service will be written in modern Java (Spring Boot preferred) and containerised for easy deployment. The work centres on embedding train-once / serve-many ML pipelines—classification, prediction, or ranking—so the code must cleanly separate model loading, inference, and API layers.
Key expectations
• Every micro-service compiles and runs locally with a simple `mvn clean package` followed by `docker build`.
• REST endpoints are documented with OpenAPI.
• Model artefacts are loaded at startup and can be hot-swapped without downtime.
• Unit and integration tests cover both happy paths and model-failure scenarios.
Deliverables
– Source code for each micro-service
– Dockerfiles (and optional Kubernetes manifests)
– README outlining build, run, and scaling steps
– Brief performance benchmark comparing raw model latency vs. service latency
Acceptance criteria
A pull-request that passes CI, returns deterministic predictions against a provided test set, and demonstrates horizontal scalability through concurrent load tests will mark the work complete.
Key expectations
• Every micro-service compiles and runs locally with a simple `mvn clean package` followed by `docker build`.
• REST endpoints are documented with OpenAPI.
• Model artefacts are loaded at startup and can be hot-swapped without downtime.
• Unit and integration tests cover both happy paths and model-failure scenarios.
Deliverables
– Source code for each micro-service
– Dockerfiles (and optional Kubernetes manifests)
– README outlining build, run, and scaling steps
– Brief performance benchmark comparing raw model latency vs. service latency
Acceptance criteria
A pull-request that passes CI, returns deterministic predictions against a provided test set, and demonstrates horizontal scalability through concurrent load tests will mark the work complete.