Authentication Service

Built using DRF, the authentication service managed user registration, login, password resets, and user roles. JWT (JSON Web Tokens) were used for secure and stateless authentication, enabling seamless session management across the microservices.

Coding Problem Design Service

This service allowed administrators and content creators to define coding problems, categorize them by difficulty, and provide sample input/output data. DRF’s flexible serialization helped create, update, and manage problem metadata efficiently.

Code Execution Service

The core feature of the platform, this service allowed users to execute code in multiple programming languages in a sandboxed environment. Code execution was handled by isolated Docker containers for security, with the service providing results such as test case success or failure.

Assessment Service

This service facilitated the creation and management of coding assessments, assigning problems to candidates, and tracking their performance in real-time. The service also recorded test submissions and graded them automatically.

Media Service (Video Interview Recording)

Leveraging AWS Chime SDK, this service enabled real-time video conferencing between candidates and interviewers. It supported recording and storing video/audio streams for future review. AWS Chime handled the video processing while the backend managed interview session metadata and file storage.

Consistency

Each service (authentication, problem design, code execution, assessment, and media) was containerized using Docker.
Docker allowed the team to manage dependencies for each service independently, ensuring that updates to one service didn’t affect others.

AWS EKS (Elastic Kubernetes Service)

All microservices were deployed on AWS EKS, which provided the scalability and resilience required for the platform. Kubernetes allowed automatic scaling of the code execution and assessment services based on traffic loads, particularly during high-traffic periods like university recruiting seasons.

Helm for Deployment

Helm charts were used to manage Kubernetes deployments. Each service had its own Helm chart, enabling streamlined updates and rollbacks. Helm also facilitated the management of configurations, secrets, and environment variables.

Persistent Storage and Load Balancing

AWS EKS provided built-in load balancing and persistent storage for databases (managed via AWS RDS for PostgreSQL) and media files (stored in AWS S3).

Real-Time Alerts and Metrics

Prometheus was integrated with the Kubernetes cluster to collect real-time metrics on CPU usage, memory consumption, request latency, and service health across the microservices.
Grafana dashboards were used for real-time visual monitoring of service performance, helping the DevOps team quickly identify and resolve performance bottlenecks or service disruptions.
Alerts were configured for key metrics such as high memory usage in the code execution service, ensuring prompt action could be taken before issues escalated.

Centralized Logging

Loki was used to aggregate and manage logs across the entire platform. Logs from all services (e.g., user actions, code submissions, API errors, and system events) were streamed to Loki, where they could be searched and analyzed in real time.
Centralized logging allowed the team to debug issues quickly, trace specific requests, and identify patterns in errors or performance issues.

Good Documentation

All REST API services followed the JSON:API specification to ensure consistency and compatibility across endpoints, with predictable request and response structures.
The API endpoints were fully documented using Swagger, which automatically generated API documentation that developers could use to test and interact with the services. Swagger’s UI provided easy access to all endpoints, including authentication, problem management, and assessment workflows.