LPL Financial

At LPL Financial I built an AWS pipeline that pulls structured feedback out of advisor call transcripts. It uses token-aware batching and multithreading, which cut processing time by 90% and let the pipeline handle far more calls.

I also fine-tuned a DistilBERT model on labeled advisor feedback to automatically sort it into categories, work that had previously been done by hand. That included building a SageMaker pipeline to handle preprocessing, training, and evaluation.