Auto DL

The process of using Deep Learning involves various steps:

• Data Acquisition
• Feature Engineering
• Model Architecture Design
• Selection of Hyperparameters
• Metrics of evaluation, Loss/Reward function
• Solving Problems that occur during training

There is a need for a platform that provides an easy interface to ease the iterations one has to take before getting a deployment-ready solution is needed.

Our Solution: - The user can make a customized DL model without writing its code, using the GUI interface - It will contain the different components or layers of the DL model - The DLMML or “Deep Learning Model Markup Language” part will convert it to an intermediate representation (IR) - This will be then passed to a parser which will convert the IR to code.

If the user wants to use Auto-DL, similar to the first step, he can input the task and other abstract inputs like “ data directory ” and the model architecture will also be generated by Auto-DL, it will truly be a one-click DL model development.

Features: - Scalability: Highly scalable as it can be deployed on a single compute system or a cluster of systems

• Data Security: Data is stored and process on premises ensuring its security, i.e. Data never leaves your machine

• Explainability: Visualize and understand what the AI model is learning

• Flexibility: Giving users choice to develop models in any framework and any programming language, thus allowing people from all backgrounds to collaborate and build solutions.