lower Fargate costs
Behravan.ai AWS compute costs reduced through cloud architecture and cost controls.
← Back to selected work
Cloud architecture · Applied AI · Delivery
Building the cloud and AI delivery system behind two SaaS products
AWS architecture, infrastructure as code, production AI workflows, security controls, and delivery pipelines for two Canada-based AI-enabled SaaS products.
- Organization
- Behravan.ai / QuickHands
- Formal role
- DevOps Engineer (AWS Architecture, IaC and Security)
- Period
- Jan 2024 – Sep 2025
- Engagement
- Contract; remote; shared engineering team across both products
Verified outcomes
Measured outcomes for Behravan.ai / QuickHands
lower Fargate costs
QuickHands AWS compute costs reduced through cloud architecture and cost controls.
release time
Behravan.ai releases accelerated with automated tests and zero-downtime delivery patterns.
01
Context
The challenge
Complexity in its real operating context.
Two AI-enabled SaaS products needed repeatable AWS environments, reliable delivery, controlled cloud spend, security controls, and production AI capabilities tied to real product workflows.
02
Accountability
The role
Where ownership sat.
As DevOps Engineer (AWS Architecture, IaC and Security), Hatef owned AWS architecture and infrastructure as code across development, staging, and production, alongside architecture and production delivery for the documented AI workflows.
03 · Architecture decisions
Decisions over decoration.
Only decisions documented in the public résumé are included here.
Codify the platform
Provisioned environments with CloudFormation and Terraform across ECS Fargate, VPC, Aurora PostgreSQL, CloudWatch, and Route 53.
Connect AI to product context
Delivered QuickHands workflows in which OpenAI and Amazon Bedrock identified home-service issues from customer images, plus an OpenAI assistant driven by product- and job-specific prompts.
Turn session content into follow-up
Directed production delivery of a Gemini workflow at Behravan.ai that processed Google Meet session content into client-facing summaries and structured follow-up points.
Treat access and delivery as architecture
Established least-privilege controls with IAM Identity Center, SSO/MFA, RBAC, AWS WAF, KMS, and Secrets Manager, and built delivery pipelines for six services.
04 · Outcomes
What changed in production.
- Reduced Fargate costs by 40% at Behravan.ai.
- Reduced Fargate costs by 30% at QuickHands.
- Reduced Behravan.ai release time from two hours to 15 minutes with automated tests and zero-downtime patterns.
- Converted incident findings into monitoring, security, delivery, and architecture improvements.
System components
- AWS
- CloudFormation
- Terraform
- ECS Fargate
- VPC
- Aurora PostgreSQL
- CloudWatch
- Route 53
- OpenAI
- Amazon Bedrock
- Gemini
- Google Maps APIs