Jacob Cutter
About Jacob Cutter
Jacob Cutter is a data scientist with extensive experience in developing and optimizing automatic speech recognition (ASR) models. He has worked at Deepgram and Insight Data Science, contributing to product development and research in natural language understanding.
Work at Deepgram
Jacob Cutter has been working at Deepgram as a Data Scientist since 2020. In this role, he has trained numerous end-to-end Automatic Speech Recognition (ASR) models using PyTorch frameworks. He has collaborated with the Engineering team to deploy these models at scale within Deepgram's next-generation production system. Additionally, he served as Team Lead of Product Development from 2022 to 2023, where he devised and reported key performance indicators (KPIs) for core speech products and evaluated next-generation deep learning architectures to enhance ASR model accuracy and performance.
Education and Expertise
Jacob Cutter earned a Bachelor of Science (B.S.) in Physics from the University of California, Davis, from 2010 to 2014. He continued his education at the same institution, achieving a Doctor of Philosophy (PhD) in Physics from 2014 to 2020. His academic background has provided him with a strong foundation in data science and machine learning, which he has applied throughout his professional career.
Background in Research and Teaching
Jacob Cutter has a significant background in research and teaching at the University of California, Davis. He worked as a Graduate Student Researcher from 2014 to 2020, where he helped build Natural Language Understanding (NLU) models, including text summarization, and deployed them as Dockerized Flask applications. He also served as a Graduate Teaching Assistant from 2014 to 2016 and as an Undergraduate Researcher from 2012 to 2014, where he sourced and curated terabytes of unstructured speech data for various research projects.
Technical Contributions
In his roles, Jacob Cutter has made technical contributions that include mining enterprise PostgreSQL databases to construct representative training and evaluation datasets. He has also spearheaded efforts to improve ASR punctuation by developing multilingual pipelines for text cleaning, data preparation, and experimental modeling using PyTorch. His work has focused on optimizing ASR models and enhancing their performance in real-world applications.