• Achieved 89% accuracy in generating labels for supervised learning using Weak Supervision, corrected label errors & created clean labels
• Benchmarked & assessed 4 HuggingFace variations of the DALL-E model for Text2Image generation, to create images within 5-30 seconds
• Built & deployed AutoML module for NLP text classification: data profiling, preprocessing, vectorization, and model flow, on AWS Marketplace
• Replicated, inferred, and explored the generation of Text2Motion, from an open-sourced pre-trained model using varied user-defined text

intern_report.mp4