01 · Platform Rebuild

Rebuilding an HR platform for scale, without stopping it

HeadHuntersHQ · Technical Lead, team of 4 · 2024-present

300K+ job postings served

100K+ applications processed

-25% projected hosting cost

<1 wk release cycle (was 3-5)

The problem

The core platform ran on always-on EC2: costs scaled with idle time, deployments were manual, and major features took 3-5 weeks to ship. A new enterprise deployment with government-grade compliance requirements was coming, and the proposed architecture would have priced it out.

What I decided

Rebuilt the core on Vue 3 + FastAPI + PostgreSQL, deployed on AWS Fargate/ECS — compute on demand, not on idle.
Redesigned the proposed enterprise architecture before launch, cutting projected hosting cost 25% while meeting compliance.
Moved releases to modular system design + agentic AI-assisted workflows, pulling the cycle from 3-5 weeks to under 1 and production incidents down 30%.
Designed PostgreSQL backup/DR with point-in-time recovery — a documented, tested restore path, not a policy document.

Decision	Why
Fargate over smaller EC2	Pay-for-use compute; idle cost drops to ~zero
Cert-based SSH over static keys	Short-lived credentials, machine identity, centralized audit
Modular services over monolith rewrite	Ship value weekly while migrating, not after
Automated PITR backup	Recovery path tested, not assumed

Vue 3frontend

FastAPIservices

PostgreSQLdata + PITR

ECS / Fargateon-demand compute

Cert-based accessaudit + DR

Outcome: platform serves 300K+ postings and 100K+ applications on demand-scaled compute; release cycle under a week; incidents down 30%; enterprise deployment passed compliance at 25% lower projected hosting cost.