The problem Quick-Serve solves

Jupyter Notebooks are one of the most popular tools to get started with building machine learning models. While they are pretty straightforward to get started with, ML models aren't the easiest thing to deploy. While fundamentally a different tool than standard python scripts, added complexity is figuring out the backend to run the inference engine, design and build the frontend UI, connect the different moving parts, all in all, a pretty daunting task for a beginner and a cumbersome process for an experienced practitioner.

We propose a tool that abstracts all these. Within a few lines of code, deploy your ML models in form of websites ,apps and cloud platforms like heroku , all in < 5 lines of code.
This tool tends to bridge the gap between Machine Learning and deployment of ML Models - often one is adept in either but almost never in both.
For beginners, who are just starting out with ML, the biggest hurdle that they face is deploying a ML model on a platform. The accuracy or precision of the model is just a number for them. With our tool they will be able to deploy their model in less than 5 lines of code and test their model in the real world and realise that these numbers are not just numbers
This tool also serves the pro ML practitioners who are not adept in web dev and/or appdev and/or PaaS deployment, for this tool automates and abstracts the deployment process so that they get on with building the model itself!

Challenges we ran into

1. Deciding on the Techstack
2. Designing the project to make it extensible
3. Figuring out the app vertical: We had to make a design choice based on user ease. Building the model from scratch would require a lot of deps on the user's end. As a workaround, we exposed the API as a public endpoint and the app makes POST requests there. Thus reducing model/computational costs and makes it far simple to prototype and quickly use.