Yusuf Mesbah

Built a dropshipping system to collect product updates from different sellers and list them on Walmart and eBay, with daily automatic updates, a script run on a cloud function google cloud platform, and storing data in Firebase.

Using Qt and OpenCV, I developed a system to automate video and image processing for a marketing company. Making templates and pipelines for faster rendering made the system accessible to inexperienced personnel.

The primary goal was to create a Python-based platform that not only interfaces with various GPT tools but also facilitates seamless data collection, GPT model training, and interaction with the trained model.
I created a user-friendly web interface, ensuring clients could input their data smoothly. Post this, the data would be used to train the ChatGPT model, allowing clients to ask questions and receive insightful responses. A key feature I integrated was the data storage mechanism that preserved valuable user interactions, paired with the capability to export this data in both CSV and Excel formats. This project encapsulated my expertise in Python development, GPT tools integration, and user experience design, resulting in a tool that met the client's expectations and requirements.

A Python application that integrates with the client's database of text data.
The goal was to filter the database based on some criteria using a web application, then summarize and group the text using information retrieval techniques and ChatGPT API so a human could read it.

Leveraged my full-stack development expertise to build a React and FastAPI application designed to streamline the navigation of administrative bureaucracy. My primary contribution involved collecting data from federal registry websites and structuring this data within a data lake for semantic search and Q&A functionalities.

This application serves as an interface for a GPT-based model to summarize complex bureaucratic documents, providing users with clear and concise answers. The back-end architecture, which I crafted, not only supports robust data interactions but also scales to accommodate the AI model's requirements. With a focus on back-end development, I established a solid data architecture that enables effective model training and fine-tuning. This back-end system is the backbone that supports a user-friendly front end, allowing for seamless interrogation of the data and presentation of nuanced answers.

In a transformative project for a client, I played a key role in transitioning established Excel-based financial models to an efficient, web-based tool using React and FastAPI. My expertise in Python was instrumental in setting up robust data processing structures and implementing complex calculations for incoming data files.

The newly created system facilitates seamless data uploads and analytics, fully hosted on Google Cloud, which has significantly enhanced the client's operational efficiency through automated data flows. I was also responsible for defining user interactions for data upload, setting criteria for data analysis, and ensuring the smooth operation of the database management system.

This project was not only about translating Excel models to a web format but also about laying the groundwork for future expansion using advanced data science methodologies. My proficiency in Python and familiarity with cloud platforms enabled me to deliver a solution that provides the client with an intuitive interface for complex data analysis tasks.

I architected a CNN-based model for classifying pneumonia from chest X-rays, employing VGG11 for feature extraction and PCA for dimensionality reduction on tabular datasets, enhancing image processing and diagnostic precision in medical AI applications.