z Ziad Khaled
ls ./selected-work --count=4

Selected work

Four real builds behind the claim on the home page — proof, not a pitch.

01

FixZone

// vehicle damage detection, shipped inside a real Flutter app

Bounding boxes kept missing small dents and scratches, so I moved from YOLOv8 to YOLOv11m-seg for real outlines, retrained it around severity levels with EGP-based repair estimates, and wrapped it in an app with an offline chatbot and a PDF report generator.

→ photo in, damage report, shop suggestions, and a PDF out, end to end.

02

JARVIS

// a multi-LLM agent platform that controls my computer and browses to learn

Built a multi-LLM router plus computer-control and web-browsing tools. Caught it once reporting "done" on a task with no file ever created — I'm actively building an output-verification layer so it can't claim success it didn't earn.

→ real, unsolved problem I'm working through — not a finished demo.

03

FlowLifeOS

// a gym-tracking app, rebuilt native so it works with no internet

Started as a WebView wrapper, which meant it needed a connection to work — exactly when a gym app shouldn't fail. Rebuilt fully native in Kotlin and Jetpack Compose so it works offline, every time.

→ running on my own phone right now — I use it.

04

BCI / EEG Motor-Imagery Classifier

// graduation research project, not a shipped product

Built a testing pipeline on the Competition IV EEG benchmark and compared CSP-LDA against a Riemannian-manifold approach for classifying left-hand vs. right-hand motor imagery straight from brain signal.

→ Riemannian manifold reached 97.92% vs. 86.11% for CSP-LDA — a strong result on a clean benchmark, not a claim about messy real-world signal.

What's coming next

This site updates as the work does.

Future posts & capstone work. I'm partway through an 8-week AI Fluency track at FlyRank — this space is where the write-ups, the agents I build, and the final capstone project land as they're finished, rather than being announced after the fact. A FlyRank completion badge goes here too once the capstone is approved.

Want to talk about a machine learning role?

Send a message straight to my inbox — I read every one myself.

Get in touch