Synthetic Data Generator & Data Janitor

Introduction

Every businesses or AI/ML models are highly depend on the clean and large amount of training data. As the data scientists we spend almost ~80% our time for data cleaning and there is absolutely no way to generate synthetic data for a specific domain or industry.

NVIDIA Jetson Orin Developer Kit and Jetson Generative AI Lab are the greatest tools ever exists for the generative AI for edge. So I can not come up with the problem where other products in the market can easily solve. Therefore I am going to build a model which runs on NVIDIA Jetson Orin Developer Kit can clean data in any data pipe lines using pre-trained transformers (I am planning to use unsupervised learning to build a foundational model and using RLHF to fine tune the model to work with specific domain or industry for data cleaning task). The other problem we face in data science or data analysis is the lack of training data, for that I am going to build another model which runs on NVIDIA Jetson Orin Developer Kit and generate synthetic data for the specific domain or industry using the very limited existing data.

The two generative AI models I am going to build and train are running on NVIDIA Jetson Orin Developer Kit. Also data collection, training and inference are performed by NVIDIA Jetson Orin Developer Kit simultaneously. Eg: You are planning to build a anomaly detection model for CNG compressors, for that you need to collect compressor vitals (such as temperature, vibration, pressure, load current and etc...) for long period of time. What if you can use my pre-trained models running on NVIDIA Jetson Orin Developer Kit which can collect, clean data (for less period of time) than generate synthetic data using the collected data and generate anomaly detection model for the compressors.