Refaktr

Build apps with AI for the real world.

Example projects and practical lessons for making AI-built software work in production.

Explore projects Read articles

Production engineering

What happens after the vibe‑coded prototype starts working?

AI can build the MVP quickly. Production engineering starts when real users, data, failures, permissions, and costs enter the system.

Security

Authorization, tenant isolation, secrets, least privilege.

Reliability

Retries, idempotency, failure handling, recovery.

Observability

Logs, metrics, traces, evaluations.

Performance

Latency, concurrency, scaling, load.

Cost

Infrastructure and inference economics.

Articles

Practical lessons

· 7 min read

I gave Claude Code the same AWS tasks before and after Agent Toolkit

Three AWS tasks, run five times each with and without Agent Toolkit. What changed — and what still required production engineering.

Read article →Draft

View all articles

Projects

Example solutions

Prodify

Take a Lovable app to AWS. Converts the project into a CloudFormation template for S3 + CloudFront hosting in infrastructure you control.

View project →

Watchy

SaaS status monitoring in Amazon CloudWatch — Slack and GitHub incidents, metrics, and alarms in your own AWS account.

View project →
Bennett Borofka

About

Bennett Borofka

Cloud architect and builder exploring production engineering for AI-built software.

Refaktr is my engineering lab for testing what it takes to make software built with AI reliable, secure, observable, and operable in production.