Lessons from Building a Lightweight PaaS

Modern Platform-as-a-Service (PaaS) solutions make application deployment easier, but they often come with trade-offs. During the process of building Velocity PaaS—a lightweight self-hosted deployment platform for internal developer teams—we evaluated several popular open-source solutions. Some were too feature-rich for our use case, while others lacked the flexibility we needed. Rather than forcing our workflows to fit existing platforms, we decided to build one tailored to our requirements.

This session shares that engineering journey from idea to implementation. Instead of focusing on Kubernetes or large-scale infrastructure, we’ll explore how a lightweight platform built using Docker, reverse proxies, SSH, Git-based deployments, and AI-assisted development can provide an excellent developer experience for small teams and internal platforms.

A significant part of the journey involved AI-assisted engineering. Rather than replacing software engineering fundamentals, AI accelerated experimentation, generated boilerplate, and helped iterate rapidly while human judgment remained essential for architecture, debugging, security, and operational decisions.

Attendees will leave with a practical understanding of when building a custom internal platform makes sense, how to avoid common pitfalls, and how AI can become a productive collaborator in platform engineering without becoming a crutch.

1. Identify common mistakes encountered while building a PaaS from scratch and practical ways to avoid them.

2. Leave with ideas and patterns that can be applied to build lightweight internal platforms for developer teams.

3. Discover where AI accelerated development—and where engineering expertise remained indispensable.