Hi, I'm Adli Qanadilo.

Converted 21 hand landmarks into 57 geometric features, including normalized coordinates, fingertip distances, angles, and extension flags. Trained a scaled RBF-kernel SVM with an 80/20 stratified split; OpenCV and MediaPipe support live webcam inference. Evaluation code produces a classification report and confusion matrix.

Represented 765 unique base game states with unordered maps, accounting for board rotations and reflections. Trained over thousands of episodes with decreasing exploration, recursive reward propagation, and Bellman value updates.

Implemented model training using gradient-ascent optimization and evaluated training and testing performance as part of machine learning coursework.

Built and tuned a neural network for digit recognition, using iterative evaluation to improve classification accuracy.

Combined sentiment and market features with feature selection, robust scaling, and configurable logistic regression, random forest, gradient boosting, and voting models. The code includes time-series evaluation and an inference pipeline that reloads the trained model and preprocessing artifacts. Evaluates F1 and ROC AUC; no unverified score is claimed here.

Built around a Next.js interface and FastAPI backend with MongoDB storage. The supplied code includes employee availability updates, schedule generation, shift swaps, and a chat-based scheduling assistant. The résumé also describes constraint-based scheduling with Google OR-Tools.

A Next.js frontend communicates with a Django REST API to manage guest profiles, preferences, and booking histories. The supplied project includes role-specific guest views and AI-generated hospitality insights using Groq. Built collaboratively during a hackathon.