At DTU Wind and Energy Systems, I implemented machine learning and hierarchical clustering algorithms to detect and correct phase imbalance in smart meter data. I scaled the algorithms to handle about 50,000 meters on up to 16 nodes on the DTU Sophia HPC Cluster.
At Brookhaven National Lab, I worked to improve a machine learning model for CPU performance prediction and implemented and benchmarked LSTM-based architectures. I proposed a hierarchical LSTM design that increased model accuracy by about 2%.
At BioMLSP Lab, I implemented early fusion models for multi-omics colorectal datasets and found that combining UNI with an autoencoder produced the best results, with a concordance index of 0.77. At Tekscend Photomask, I developed a web application that summarized real-time tool data across production stages and pinpointed bottlenecks.
