Kevin C.

I build production web apps, and the systems that catch problems before customers do. Since May 2026 I've worked as a developer and QA engineer for Vidonary, an AI video creation platform built on Next.js, AWS, PostgreSQL and Redis. I've delivered more than 120 tasks there, spanning features, bug fixes, infrastructure and monitoring. RESULTS • Catch about 83% of bugs in new code before they reach production (July to mid-October, 78–90% month to month). That counts only bugs caught by my own testing and automation, out of every production bug traced back to a recent change. • Built an AI QA bot that runs alongside my manual testing. In its first full week it wrote 51 test reports and filed 20 bug tickets. • Fix what I find instead of just filing tickets: 191 pull requests merged since July, on top of testing. WHAT I'VE DELIVERED • Features: shipped 23 user-facing features, including switching between screen and camera mid-recording, word-by-word highlighting in a text-based video editor, bulk download and delete, and subscription plan gating. • AI video backgrounds: made an AI background-replacement feature (blur, colour or a custom image behind the speaker) fast and production-quality. Added custom image upload with validation and tests, raised the segmentation model's resolution 4× so people no longer looked see-through, and fixed distorted portrait and rotated phone clips. Cut processing time for a minute of 1080p video by a third (189 s to 126 s on the same server), with CPU throttling down from 99% to 9%. • Bug fixes: fixed about 30 bugs across the video editor, the AI editing tools and the export pipeline, including failed exports on large and high-resolution files. • Production monitoring: took production from one alert to full coverage: error tracking on every service, AWS alarms across the stack, heartbeat checks for scheduled jobs, alerts when key workflows like exports or payments start failing, and alerts that reliably reach the team without flooding them. • Automated end-to-end testing: built and maintain Playwright tests that sign up as a new user and walk through the core product on every deploy, and hourly in production. They've caught regressions before customers saw them. • AI quality checks: built an automated test where an LLM scores AI-generated scripts against a quality bar, then used it to fix off-topic output after a customer complaint. • Performance: fixed autoscaling so AI jobs no longer queued up under load. • Integrations: fixed scheduled publishing to Instagram, Facebook, YouTube and TikTok. • Data repair: corrected 1,374 videos recorded at the wrong orientation, using a repair job that was tested safely first and ran without slowing the live app. • Reporting: set up automated weekly reports on infrastructure cost and feature usage. • Release management: own the path from staging to production. I test every change before release and ship batched releases of 30+ updates at a time. PERSONAL PROJECT The Chronicle, a multiplayer web app where an AI runs a tabletop role-playing game for a group of players (Node/Express, Firebase, Anthropic API). I'm comfortable stepping into an existing codebase, learning how it works, and improving it without breaking what already works. STACK React, Next.js, TypeScript, Node.js, FastAPI, PostgreSQL, Redis, AWS (Lambda, EKS, CloudWatch), Playwright, Sentry, GitHub Actions, Stripe, LLM integration (Claude, OpenAI)