Peter Szemraj
Statistics MSc - ETH Zürich
About
Statistics Student at ETH Zürich. A dual Polish and US Citizen.Three years of prior work experience in the US Energy Sector: supply chain operations, project management, and analytics.Interested in applications of NLP, computer vision, and unsupervised learning to biochemistry, written document information retrieval, and geospatial topics.Contact info below, and my calendar booking tool can be found here.
Projects
Generalizable Summarization Models - A case study on creating NLP models that summarize day-to-day text effectively with LED & the BookSum dataset.
BoulderSpot - extended MVP from CASSINI Hackathon to search for new bouldering (rock climbing) outdoors with a CNN trained on Swiss public satellite data.
BoulderAreaDetector: A computer vision deep learning app that finds new rock climbing and bouldering areas in satellite images.
vid2cleantxt: a transformer-based pipeline for turning heavily speech-based video files into clean, readable text from the audio.
Oral Wound Healing Modelling, analysis of time series patient-treatment panel data in R from a project in the Seminar for Statistics.
Hackathons
Part of the team placing 3rd in the CASSINI EU Space4Impact Hackathon in June 2021 with NatureGeoDiscoverer, a remote sensing "area of interest" PoC
Publications
Wu, J., Zodrow, K. R., Szemraj, P. B., & Li, Q. (2017). Photothermal nanocomposite membranes for direct solar membrane distillation. Journal of Materials Chemistry A, 5(45), 23712-23719.
Metcalf, A., Szemraj, P., Arredondo, J. H., & Worstell, J. H. (2017, March). Distillation: An Under-Plowed Field. In 2017 Spring Meeting and 13th Global Congress on Process Safety. AIChE.
Education
| Institution | Degree | Date |
|---|---|---|
| ETH Zürich | MSc Statistics | Exp. Fall 2023 |
| Rice University | B.S. in Chemical and Biomolecular Engineering | May 2017 |
| Glenbrook North High School | High School Diploma | May 2013 |
Contact
Portfolio & Other
GPT-Peter: Language Model
Is a generative language model of Peter possible?
Telegram
Login or create a telegram account to message the bot.Responses will be computed on a local Python runtime and sent back to your account on Telegram.