Jonathan Nafziger

Programmer, Physicist, Problem Solver

Experienced Software Engineer with a Physics PhD and a passion for solving problems, playing with software, and learning new things.

Contact me PDF version

Lead Python developer for NLP, plagiarism detection, AI content detection, and LLM-based chat assistant services.
Improved plagiarism check times from a one minute average down to 15 seconds.
Architected a paraphrase detector using sentence similarity models and dynamic programming.
Built an event processing system handling 5k events/s with novel visualizations of the student writing process.
Developed NLP algorithms to generate actionable feedback on student writing.
Migrated AI detection and ML grammar correction to a dedicated GPU server, achieving a 20x speedup.
Built and deployed new Python services based on PyTorch, FastAPI, Helm, and Kubernetes.
Transitioned Analytics Service to use BigQuery to resolve issues with flaking data pipelines.

Built and managed data collection and labeling platforms using Django, Helm, and Kubernetes on GCP.
Built a human-in-the-loop labeling tool for efficiently correcting machine-generated transcripts.
Grew user base to 500+ active users, generating 3,000+ hours of domain-specific speech data.
Designed and built dialog act and emotion recognition models.
Collaborated with ML engineers to build and deploy state-of-the-art speech recognition and TTS models.
Built and managed data collection platform Gridspace Mixer, where users call in and talk to a partner in a specified domain scenario.

Provided support and development for wordpress and drupal sites using html, css, javascript and php.
Completed 11 Coursera courses including the Deep Learning Specialization, building foundational skills in machine learning and data science.

Developed novel computational methods within the field of density functional theory.
Wrote MATLAB software to solve problems involving ODEs and PDEs as well as corresponding inverse problems.
Published 12 peer-reviewed papers and presented at 13 domestic and international conferences.
Integrated new algorithms into existing FORTRAN software packages
Translated abstract ideas into custom MATLAB software for use in research
Implemented version control and unit tests to increase the quality of software developed by our research group

Helped to teach courses on electricity and magnetism, electronics and general physics as an undergraduate at Goshen College and graduate student at Purdue University