Back to Projects

Professional Networking Platform — Cloud & DevOps Internship at OREVIT

Project Overview

OREVIT  ·  June – September 2026  ·  Sole Cloud/DevOps engineer on the team

A multi-tenant SaaS platform connecting professionals across three business lines: recruitment, a freelance marketplace, and B2B tender management. Each client organization operates inside its own tenant boundary, sharing the underlying platform without sharing data.

I was the only Cloud/DevOps engineer on the team, which meant owning the infrastructure end to end: designing the AWS account structure, writing the Terraform codebase, building the deployment pipelines, and putting observability and cost controls in place. Application teams shipped against a platform they could deploy to themselves, without needing to touch the AWS console.

The guiding constraint was that nothing should exist only in someone's console history. Every environment is described in code, every deployment goes through a pipeline, and every production change is traceable back to a commit.

Cloud Architecture

• AWS Organizations with separate Dev, Staging and Production accounts, so environment isolation is enforced at the account boundary rather than by naming conventions or IAM policy alone.
• Terraform as the single source of truth: remote state in S3 with DynamoDB state locking, and reusable modules so a new environment is a set of variables rather than a copy-paste of the previous one.
• Multi-AZ VPC with public and private subnet tiers, application workloads kept off the public internet.
• ECS Fargate services behind an Application Load Balancer, with CloudFront in front for static delivery and edge caching.
• RDS PostgreSQL Multi-AZ, encrypted with KMS, for transactional data; ElastiCache Redis for sessions and caching; S3 for document and asset storage.

CI/CD Pipeline

• Containerized builds with Docker, orchestrated by GitHub Actions.
• Each pipeline run builds the image, runs the test suite, and performs a secret scan before anything is published.
• Images are pushed to Amazon ECR with vulnerability scanning enabled, so a known-vulnerable image is caught before it reaches a task definition.
• Deployment to ECS happens through OIDC federation: GitHub Actions assumes a scoped IAM role using a short-lived token. There are no static AWS access keys stored in the repository or in CI secrets.

AI Features

• Amazon Textract for CV parsing, turning uploaded résumés into structured candidate data.
• Amazon Bedrock for LLM-backed features, keeping model access inside the account boundary.
• Semantic search built on RDS PostgreSQL with pgvector, avoiding a separate vector database and keeping embeddings under the same backup and encryption policy as the rest of the data.

Observability & Cost Control

• CloudWatch for centralized logs, service metrics and alarms on the signals that indicate user-visible failure.
• Grafana dashboards giving the team a single view across services rather than per-service console pages.
• AWS Budgets and Cost Anomaly Detection configured from the foundation, not retrofitted after the first surprising invoice.

Technologies Used

Architecture Diagrams

Repository

The codebase is proprietary to OREVIT and is not publicly available. Architecture and implementation details can be discussed on request.