Hello world, I'm

Computer Science student at York University, focused on AI and machine learning. I build models that find patterns, make predictions, and turn raw data into decisions worth acting on.


Software Engineering Intern

Jan 2026 – May 2026

Treepz • Toronto, ON

→ Rebuilt how the company website publishes content: updates went from 2–3 days to under 10 minutes.

  • Architected a type-safe API service layer spanning 10 files for a Strapi CMS migration, building a complete TypeScript interface system, 7+ high-level API methods (post retrieval, filtering, pagination, static generation support), and standardized response contracts with type guards that eliminated an entire class of null-reference errors at compile time
  • Designed data transformation mappers to flatten Strapi's deeply nested response structure into clean application models, implementing fallback chains for missing fields to prevent malformed data from ever reaching the UI
  • Implemented Static Site Generation with Incremental Static Regeneration (revalidating every 3600 seconds), replacing a manual JSON-based publishing workflow and cutting content publishing time from roughly 2-3 days down to under 10 minutes
  • Added dynamic SEO infrastructure including JSON-LD structured data, a programmatically generated sitemap, and per-page metadata
  • Collaborated with a fellow intern on a shared dashboard search feature, owning backend filtering logic and API contract
  • Followed a feature-branch Git workflow, submitting weekly pull requests for senior engineer review and incorporating feedback before merge

Lung Cancer Imaging Pipeline

An AI model for spotting lung cancer looked 99.96% accurate because it was being tested on near-copies of images it had already seen. I caught it and rebuilt the testing so the score is honest.

→ Honest accuracy: 99.30%, reported with its known limits · 228 automated tests

Python · Deep Learning · Medical Imaging · Docker

[ code ]

UniQ — University Policy RAG System

Ask a question about York University policy and get an answer that cites the exact document it came from, or an honest "I don't know" when the rules don't say.

→ Correctly declines 92% of questions it can't answer · runs on a laptop for $0

Python · RAG · ChromaDB · BM25 · Evaluation · LLM

[ demo ↗ ][ code ]

Distributed Media Processing Platform

Upload a video and get back a transcript, captions, and highlight clips, and the system keeps going even when servers crash mid-job.

→ Processes about 12 minutes of video per minute on a laptop · recovers when a server crashes mid-job

Python · Celery · Redis · PostgreSQL · Kubernetes · Distributed Systems

[ code ]

More projects

RouteSmith — LLM Query Router

Decides whether a cheap AI model can handle a question before paying for an expensive one.

→ Predicts when the cheap model will fall short, and shows which kinds of questions trip it up

[ code ]

pydeadweight

Finds software packages a Python project installs but never uses, so they can be safely removed.

→ Read-only by default, so it never deletes anything without asking

[ code ]

Facial Expression Recognition

Reads your facial expression through a webcam and names the emotion, live.

→ 7 emotions in real time · no internet connection needed

[ code ]

Hand Tracker

Draw in the air with your finger. Any webcam becomes a sketchpad.

→ Tracks 21 points on your hand in real time · no touchscreen needed

[ code ]

Black & White Image Colorizer

Upload an old black-and-white photo and get it back in color.

→ A full web app, from the upload page to the server that colors the photo · work in progress

[ code ]
Baybars Al-Zibdeh, portrait rendered in amber phosphor duotone

I'm Baybars, a Computer Science student at York University. I build machine-learning systems in Python, and the problems I like most are the ones where the data works against you: data recorded over time that quietly leaks the answer into the test, or categories where one outcome is far rarer than the rest. Those are the cases where a model can look great on paper and still fail in the real world.

From January to May 2026 I was a Software Engineering intern at Treepz in Toronto. I moved the company website onto a new content system, built the code that connects the two, and changed how pages are generated, so updates that used to take two or three days now go live in under ten minutes. Outside work I build things that see: an app (in progress) that restores color to black-and-white photos, one that reads emotion from faces, and one that turns a webcam into a sketchpad. I've worked through machine-learning and generative-AI coursework from IBM and Databricks. Next, I want to work on problems where a model stops being a demo and starts making real decisions.

CONTACT

Baybars Al-Zibdeh

Certifications

Machine Learning with Python (IBM) · IBM Z Xplore (IBM) · Fundamentals of Encryption & Quantum-Safe (IBM) · Generative AI Fundamentals (Databricks) · Databricks Fundamentals (Databricks) · Scientific Computing with Python (FreeCodeCamp) · Foundational C# with Microsoft (Microsoft) · Claude Code in Action (Anthropic Education)