Muhammad Khubaib Raza

Working on this project was a great learning experience, ranging from room item detection similar to Airbnb amenity detection to text generation (table to text).

The apartment photo database includes the price of the apartment, floor, latitude, longitude, nearby places, and whether it has parking. First, we had to eliminate duplicate photos from one apartment, so we would not detect items repeatedly. In the second step, the photos are then classified into room types. In the third step, count the number of bedrooms, baths, and other rooms, as well as detect the amenities. The fourth step is to create a dataset for description generation, and last but not least, to generate the description from the transformer model.

A great learning experience that involved integrating ChatGPT API and Twilio to develop an automated voice-based customer support system.

The system's functionality enables customers to call a business's Twilio phone number and ask a query, which ChatGPT then processes to provide personalized responses based on the customer's input. The ChatGPT model has been trained to understand and respond to a wide range of customer queries.

Developed an OCR (Optical Character Recognition) system using the Paddle OCR framework to recognize and extract text from invoices accurately. The SageMaker Pipeline was used to deploy the system, which enabled continuous model training and deployment with built-in model monitoring and version control. The system was implemented to automate invoice processing for a small business, significantly reducing manual data entry time.