One of the most significant challenges encountered during the project's development revolved around hosting and utilizing our text-to-image model. Initially developed on Jupyter Notebook locally, we quickly realized its limitations for testing and production purposes. Our research led us to explore cloud platforms like AWS and Google Cloud, renowned for their backend model deployment capabilities. Additionally, we discovered Intel's Cloud Developer platform at a career fair, boasting access to cutting-edge GPU chips.

Upon deciding to leverage the Intel Cloud Developer platform, we encountered hurdles in configuring and testing our model due to our team's lack of expertise in cloud infrastructure. Fortunately, with the assistance of Rahul, an Intel volunteer, we overcame these obstacles and successfully deployed our powerful model.

However, we faced another challenge in establishing communication with our front end, as the cloud instances lacked static IPs. After weighing various solutions, including implementing a jump host server or utilizing temporary hosting with NGROK, we opted for Google Firebase. Its user-friendly interface and file storage capabilities made it an ideal choice.

Throughout this journey, not only did we accomplish the deployment of our instance, but we also gained invaluable insights into the capabilities of Intel GPUs offered by the Intel Developer Cloud. This experience underscored the importance of adaptability and collaborative problem-solving in overcoming complex technical challenges.