crafting lean software and resilient systems.

PromptVibely was built out of frustration with messy, unstructured text prompts in early AI workflows. It offered developers an intelligent editor to compose, test, and export prompt templates into strictly validated JSON schemas. Users could manage dynamic variables, version control prompt diffs, and organize templates into dedicated team workspaces.

Within the first 30 days of public launch, PromptVibely exploded to 1,853 registered accounts. But instead of doubling down on marketing, user acquisition, and sales, I fell directly into the classic developer trap: endless technical perfectionism. I spent weeks refactoring database migrations, obsessing over microseconds of query performance, and tweaking micro animations. While the engineering was immaculate, momentum waned, teaching me the foundational rule: distribution always beats code vanity.

Devpply targeted an acute problem facing college graduates: automated applicant tracking software discarding resumes before any human recruiter ever saw them. Devpply provided algorithmic resume formatting, keywords analysis, and PDF output guaranteed to parse accurately across major applicant tracking systems in under 3 minutes.

The platform crossed 1,500 verified student accounts within its first seven days. However, running a direct-to-consumer utility demanded constant customer support, document format troubleshooting, and continuous marketing operations that pulled me away from deep systems engineering. I paused the platform to focus on building large-scale backend infrastructure and distributed systems.

Devpply is now back on its own domain, devpply.com, and I am rebuilding it as v1. The relaunch starts from the distribution lessons above: a clear first-run experience, measurable onboarding, and marketing planned alongside the code rather than after it.

Around six months ago, I designed and started implementing mini-waf — a lightweight, programmable Web Application Firewall created to sit as a protective proxy layer ahead of origin application servers. The goal was to intercept incoming HTTP payloads, inspect headers for SQL injection, cross-site scripting (XSS), and malicious user-agents, and apply strict rate-limiting policies directly at the reverse proxy stage.

While the inspection regex pipelines and local reverse proxy routing worked in isolated tests, a real WAF fundamentally requires live, edge deployment to inspect raw public internet traffic, handle DNS routing, terminate SSL/TLS certificates, and manage kernel iptables under hostile scanning. At that time, I did not have an actual, dedicated VPS server with a static public IP to deploy and run it in the wild.

Because of that exact infrastructure bottleneck, I couldn't implement mini-waf in a true production environment, and I made the deliberate decision to drop the application. But hitting that wall was a catalyst: it proved to me that writing smart application code means nothing if you don't have the physical or virtual infrastructure to run it. That failure is what drove me directly into mastering Linux server management, Oracle Cloud VPS provisioning, network routing, and reverse proxy architectures.