DevOps Intern
Klickit is an Egyptian fintech that handles fee payments for schools and universities. I spent six weeks on its DevOps team, one of two interns working under a DevOps and Cloud Engineer, mostly on Spotlight, an internal ticketing platform with an Angular frontend and a .NET API, and its user-journey SDK. Nginx, Docker, Graylog, and GitHub Actions ran underneath both, all on a single Azure VM.
Internship evaluation with excellent ratings
What I worked on
Observability and deployment
- Nginx reverse-proxy and TLS configuration in front of Spotlight's Angular frontend and .NET API
- A Filebeat pipeline shipping Nginx, backend, and systemd logs into a Dockerised Graylog stack, plus the dashboards built on top of it
- Linux administration on the Azure VM: systemd services, diagnosing memory pressure, and working through issues over SSH
- Runbooks and a technical write-up of the system for whoever picks it up next
The user-journey SDK
- klickit-user-journey-sdk, an Angular library published as an npm package that auto-captures navigation, clicks, form submissions, and unhandled errors from Spotlight
- Checking what the SDK collects: from my read of the code, its idle-session tracking only uses activity timestamps, not keystrokes or field values
- The delivery side of the pipeline: client-side batching, with retries and backoff for failed or rate-limited requests
Completed Work
Graylog observability stack
A central logging setup that ships Nginx, backend and systemd logs through Filebeat into Graylog. Two incidents came up along the way: a Graylog outage, and a Filebeat misconfiguration that was dropping logs before they reached Graylog.
- Built with
- Graylog, Filebeat, MongoDB, Docker Compose
- Contribution
- Diagnosed and fixed both incidents alongside my co-intern, then wrote them up as runbooks.
- Outcome
- A logging pipeline the team can query and alert on, plus documentation for next time either error shows up.
Spotlight's deployment path
Nginx sits in front of Spotlight's Angular frontend and .NET API on a single Azure VM, with GitHub Actions handling the build-and-deploy workflow for each side.
- Built with
- Nginx, GitHub Actions, Azure, Linux
- Contribution
- Worked through the reverse-proxy configuration and traced a couple of deployment failures back to their root cause.
- Outcome
- A documented, repeatable path from a code change to a running deployment.
User-journey SDK review
klickit-user-journey-sdk auto-tracks navigation, clicks, form submissions, and errors in Spotlight, batching events client-side before sending them on to the telemetry API.
- Built with
- Angular, TypeScript, npm
- Contribution
- Read through and documented the SDK's batching, delivery, and idle-detection logic.
- Outcome
- A written reference for how user-behaviour data moves from a browser click to storage, plus my finding, from reading the code, that the SDK does not capture keystrokes or field values.
Technical Stack
Technical skills
Nginx, Docker, Linux administration, CI/CD
Observability
Graylog, Filebeat, MongoDB
Tools and platforms
Azure, GitHub Actions, MobaXterm
Problems along the way
Graylog itself went down partway through, which meant the tool we relied on to debug problems had become the problem.
Nginx, the backend, and the Graylog stack all ran on the same VM, so a fault in one layer could take down the tool meant to diagnose it.
Worked backward through Nginx, the Graylog containers, and the VM's own memory and process state until the cause turned up.
Graylog came back up, and the incident became one of the runbooks I wrote for next time.
Filebeat was shipping to the wrong destination for a while, so logs that looked like they were reaching Graylog quietly weren't.
Traced the pipeline stage by stage (the source logs, the Filebeat config, the Graylog input) until the mismatch turned up.
A working log pipeline again, and a clearer mental model of where telemetry can silently go missing in a multi-hop system.