Every model begins as noise.
I'm Fadil — a builder working toward a career in AI/ML. I label the data, tune the models, and ship the apps. Scroll to learn more! ↓ training begins
First, the data gets labeled.
Models only learn what someone teaches them. I've drawn the bounding boxes, ranked the outputs, and tagged the datasets that models train on. It's also where I learned to look closely.
Calgary, AB
Ranked LLM outputs for benchmarking, labeled UI screenshots with 100–500+ bounding boxes per image for computer vision training, and tagged multimodal datasets to improve classification accuracy.
Calgary, AB
Ran retail popup events end-to-end — setup, sales, customer service — processed weekly order fulfillment, and audited inventory to maintain stock accuracy.
UCalgary
Attend events like a hands-on Databricks workshop (DSMLC × ADSS) covering workflow automation, data warehousing, and visualization on the Databricks platform.
Then, the model trains.
Every project below is like a training run: data in, working software out
Hotel Booking Cancellation Predictor
Built and tuned 5 ML models on 119K+ bookings to predict cancellations. Random Forest reached 0.9415 ROC-AUC after hyperparameter tuning with GridSearchCV, on top of a full preprocessing pipeline for splits and categorical encoding.
Spotify Playlist Manager
Desktop app for managing playlists via a GUI, with CSV file I/O and 30+ JUnit tests. Built on clean OOP — inheritance, abstraction, polymorphism.
IMDB-Style Movie Database
Modular movie library querying 10,000+ titles through the Watchmode API, with a recommendation quiz and local user authentication.
Tic-Tac-Toe GUI Game
2-player and AI tic-tac-toe with a live scoreboard and a custom heuristic AI for defensive blocking and move selection.
Movie Player
Swing desktop app for managing a movie library, with secure login for 10+ accounts and persistent local storage via Java file I/O.
Eventually, it converges.
Here's the current checkpoint.