sajjit.soti
← Back to projects

Case study — Personal cloud architecture project

Designing a telemedicine cloud architecture for data that can't afford to leak

Healthcare platforms carry some of the most sensitive data a system can hold, and telemedicine adds a second constraint on top: real-time video between patient and provider has to stay reliable while every layer around it stays locked down. This project designs a secure, scalable AWS architecture around that pairing — strict security boundaries and fault isolation, without sacrificing the responsiveness a live consultation needs.

WAF + Shield

Edge and perimeter protection

Cognito + MFA

Token-based identity & access

3

Isolated auto-scaling microservices

KMS

Encryption for data at rest

01

Edge and perimeter

The first layer of defense sits at the edge: AWS WAF and Shield filter malicious traffic before it reaches the application, while GuardDuty continuously watches for anomalous account and network activity. Everything sensitive at rest is encrypted through AWS KMS— encryption isn’t an add-on here, it’s a default.

02

Identity: no password-only paths

User authentication runs through Amazon Cognito with MFAenforced, using token-based sessions rather than long-lived credentials — the same principle behind the identity checks in the self-healing SOC project, applied here at the platform’s front door instead of after the fact.

03

Isolated microservices, live video

Compute is split into three separate services — Telemedicine, Admin, and Records — each on its own EC2 Auto Scalinggroup. Keeping them isolated means a problem in one (a scaling event, a misconfiguration, an incident) doesn’t cascade into the others; the Records service handling patient history doesn’t share a blast radius with the public-facing Telemedicine service. Real-time video between patient and provider runs through the Amazon Chime SDK, integrated directly into the Telemedicine service.

04

A data layer built to survive failure

Storage and state are spread across Amazon RDS, ElastiCache, and S3, backed by automated AWS backups — the goal being that no single component failure turns into data loss. Amazon SageMaker sits on top for advanced data insights, giving the platform a path to analytics without pulling raw patient data out of its secured layer.

05

What it reinforced

Designing for healthcare data sharpens a habit that matters everywhere: treating network isolation and fault boundaries as first-class design decisions, not an afterthought layered on once the architecture is “done.”