Hi, I'm Ghani Haider

I am a computer science undergrad with experience and interest in Deep Learning. I am seeking to leverage my deep learning skills to deliver insights and implement action-oriented solutions to complex problems.

Implemented Research Papers:

Paper Code	Summary

Transformer's Decoder	Implemented transformer's decoder for language generation from scratch using paper Attention is all you need to generate a sequence of characters given an input/prompt.
Vision Transformer	Implemented vision transformer (ViT) for image classification using paper An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale.

Deployed Projects:

Project	Link

Seoul Bike Rent Prediction	Web App

Projects & Open-Source Contributions:

Topic	Summary

Deep Learning
Landscape Recognition	Implemented Tiny VGG model (inspired from cnn-explainer website) on landscape images dataset from Kaggle using PyTorch and torchvision.
Generating Shakespeare	Implemented transformer based language generation model from scratch using paper Attention is all you need to generate Shakespeare's play.
2D Self-driving Simulation	Implemented Deep Q-Network (DQN) with greedy action selection policy for simulating a self-driving vehicle using OpenAI Gym car racing environment and CNN architecture.
Gender Determination by Morphometry of Eyes	A model which uses scanned images of patient’s eye and determine their gender for anthropometric analysis of the human face using Artificial Neural Network (ANN)

Data Analysis & Visualizations
Mushroom Dataset Exploration	The objective is to analyze the mushroom dataset to create a predictive model in order to classify edible and poisonous mushrooms. We also have to create a general guideline for features that are important to consider for future references.

Regression Projects
Seoul Bike Rent Prediction	Understand the trends in the data and indentify key factors affecting demand for rental bikes in Seoul as well as predict the amount of rental bikes required per hour.
Used Car Price Prediction	Using information about used cars listed on cardekho, predicting car prices based on given features and trends in the data.
Rock Density Prediction	Objective is to analyze the performance of different regression models on given dataset (use of X-rays in an attempt to know rock density)

Classification Projects
Urdu Speech Emotion Recognition (SER)	Developed a Speech Emotion Recognition (SER) model for Urdu language using SEMOUR dataset in order to correctly classify the Urdu speech and audio inputs to emotions.
Banknote Authentication	Using Random Forest Classifier for distinguishing genuine and forged banknotes based on extracted wavelet information from genuine and forged banknote-like specimens.
Heart Disease Classification	To confirm 100% if a patient has heart disease can be quite an invasive process, so creating a model that accurately predicts the likelihood of heart disease to help avoid expensive and invasive procedures.

Time Series Forecasting
Hourly Energy Consumption	Given hourly power consumption data from PJM's website (an electric transmission organization in USA), understanding both short and long terms trends in energy consumption and to predict energy consumption over a certain period.

Natural Language Processing
Text Classification using Naive Bayes	Given a dataset of textual summary of medical queries classified into five different categories, built a naïve Bayes classifier to predict these categories for future queries.
Collaborative Filtering : Matrix Factorization	Implemented collaborative filtering from scratch to make recommendations to users U for items I. Used a model based approach that applies matrix factorization to factorize a rating matrix (𝑅) into User features (𝑃) and Item features (𝑄).

Quantum AI
Quantum Artificial Neuron	Implemented the quantum information-based algorithm proposed by Tacchino et al., which is a quantum model of a binary-valued perceptron.