This is a list of interactive ML applications I built. Please open the links in new windows, since currently they won't link back.
- Extraction of features from earnings reports and calls with NLP and feature engineering
- Built an ensemble model with genetic algorithm to search hyper-parameter space for best prediction score
- Serverless compute and render using AWS Lambda
- Full stack deep learning system for text OCR (PyTorch)
- Image segmentation with Fully Convolutional Networks (92.8% pixel accuracy)
- Character prediction with CRNN model and CTC loss (19.2 Character Error Rate)
- Implemented CI pipeline, dockerization, web server to enable model iteration, monitor and serving
- Highly specific, condition-agnostic predictor for protein-DNA binding events
- Applied a sliding window to largely generate negative labels in order to maximize prediction specificity
- Serverless compute and render using AWS Lambda