Prateek Bansal

Optimized and deployed an image and text classifier to remove ~10k/week of irrelevant content from a LinkedIn feed Designed and implemented sampling method to optimize IC productivity by +2 Dev-weeks/Qtr Prompt Engineering on GPT 3.5 responses to increase classifier performance through effective model training by notable +20% Led and Implemented ETL process to increase labeled data by +235% for enhanced model development

1. Research and Implementation of different CNN and RNN based architecture for sequence data. 2. Worked on semantic segmentation of road defects using Deep Learning Architecture(HRNET) to get detailed representation and to improvise it based on different metrics. 3. Design and Implementation of MongoDB database for POC. 4. Currently working on MLP model to find the road goodness value based on features extracted from defect segmentation.

I was responsible for the enhancing, porting and testing functionality of NonStop Server Utilities. Used shell scripts to automate certain test cases, validate the changes and to enhance test speed. From late 2018, I started doing machine learning projects as a part of innovation drive, one such project is "An Efficient Based Machine learning method to analyze performance problems on NonStop".

I was tasked with the design and implementation of TOP Utility in C language for Non-Stop Platforms.

Very Deep Learning, Machine Learning, Application of machine learning, 3D computer Vision, Data Visualization, 2D image Processing, Visual Analytics. At Present - Doing thesis in DFKI and TUKL on Image Synthesis using Generative adversarial network.

Artificial Intelligence, Human Computer Interaction, Linear Algebra, Software Engineering.