Achraf Aourik
With 4+ years of experience working as a data scientist with a strong background in statistics, analytical modeling and programming, I aim to develop accurate machine learning applications and synthesizing their insights in a concise and robust story for presentation to stakeholders and clients and therefore help them achieve their business goals and increasing their profits.
2X AWS Certified : Machine Learning Specialization and Cloud Practioner
View My LinkedIn Profile
Personal projects in data science and machine learning
DocChat
This project aims to give a choice about whether the user would like to leverage ChatGPT or use an OpenSource LLM to answer the questions.
The objective is to provide these AI models with access to a rich database of internal documents and allow them to retrieve the most relevant pieces of text in response to user queries.
The system leverages advanced text retrieval and information ranking mechanisms to sift through a vast array of internal documents. Using the provided context, the system discerns user intent and identifies the most pertinent documents. These documents are then processed, extracting the most appropriate segments that can provide the user with the most accurate and comprehensive answer.
Churn detection project
Conducted extensive analysis on real-world data of Taiwan’s leading music streaming platform. In this project, I performed all of the steps in a machine learning worfklow: data cleaning, exploratory data analysis, feature engineering, modeling (supervised and clustering) and model evaluation. I was able to create a highly accurate classification model that is able to accurately predict which clients are more likely to churn.
Some select graphs of the EDA performed comparing the two profiles or churners and loyal subscribers:
Distribution of churn data relative to “auto_renew”:
Distribution of churn data relative to “registration_method”:
Clusters centroids generated after performing clustering:
2D Visualization of the clusters using PCA:
The evolution of performance after every iteration:
The Feature Importance of the final XGBoost model:
Financial Helper Bot
A chatbot that is capable of creating bank accounts, checking their balance and performing transactions on these accounts. It is also capable of doing other non-financial related tasks such as answering FAQ questions and chitchat with users.
The architecture of the project is illustrated below:
A demonstration of how the bot works is shown below:
