Building beyond the obvious.
I'm Ghania, A Computer Science student focused on AI/ML and full-stack development. I work with model pipelines, backend APIs, and modern web tech to turn ideas into systems that actually work.
A bit about how I work
I'm a Computer Science student at FAST-NUCES, currently in my 5th semester, focused on AI/ML and full-stack development.
I care about the full path a project takes — from cleaning messy data and choosing the right model, to explaining why it made a decision, to putting it behind an API or interface someone can actually use. I'm equally comfortable comparing models in a notebook and debugging a FastAPI backend.
Outside coursework, I'm interning in AI/ML at Deimos Tech and doing marketing for a healthcare startup — a good lesson in explaining technical work to a non-technical audience.
Where I've worked
Two internships running in parallel — one technical, one client-facing.
AI/ML Intern
- Built an end-to-end fraud detection pipeline handling class imbalance with SMOTE, trained with XGBoost, tracked via MLflow, and explained with SHAP
- Compared Logistic Regression vs. Decision Tree on the Breast Cancer Wisconsin dataset with cross-validation and hyperparameter tuning
- Fine-tuned a pretrained ResNet-18 on CIFAR-10, improving test accuracy from a 79.63% baseline to 91.42%
Social Media Marketing Intern
- Planned and produced content across Instagram and Facebook, including a 30-day content calendar and pre-launch campaigns
- Set up and managed platform infrastructure — Facebook Page, Instagram cross-posting, LinkedIn presence
Projects
A mix of applied ML, systems programming, and full-stack apps — picked to show range. Scroll to see all seven →
Chat With Your Documents
A retrieval-augmented chat app — upload PDFs, DOCX, or TXT files and get answers grounded strictly in that content, with page-level citations for every response.
Fraud Detection ML Pipeline
End-to-end pipeline for detecting credit card fraud on an imbalanced dataset — SMOTE resampling, XGBoost, MLflow tracking, and SHAP explainability, served through FastAPI.
CIFAR-10 Fine-Tuning
Fine-tuned ResNet-18 on CIFAR-10, lifting accuracy from 79.63% to 91.42% with a full ablation study on unfreezing depth.
Concurrent Banking System
OS-concepts simulation covering scheduling, synchronization, deadlock avoidance, IPC, and memory management.
Dynamic Pathfinding Agent
Live A* vs. GBFS visualizer with mid-run replanning when obstacles spawn on the grid.
Court Piece
A playable full-stack card game — FastAPI backend owns game logic, live 4-player browser frontend.
Station — Weather Dashboard
React + Vite dashboard with a serverless proxy so the API key never reaches the browser, plus rate limiting and caching.
Skills
BS Computer Science
Pre-Engineering (FSc.)
Let's Connect
Open to AI/ML and software development internships, and new-grad roles. I reply fastest over email.
Say Hello →