Kyle Lo Email & Phone Number

Seattle, Wa, Us

Greater Seattle Area

• I am co-leading the open dataset research effort for OLMo, which culminated in the release of Dolma, a large, diverse open dataset for language model pretraining. Both OLMo and Dolma papers won best paper awards at ACL 2024 🏆. I've also co-led the creation of scientific literature corpora, including S2ORC, the largest, machine-readable collection of open-access full-text papers to-date, and CORD-19, the most comprehensive, continually-updated set of COVID-19 literature at the time.• I work on specializing language models to knowledge-intensive domains, like science. I was a developer of SciBERT, one of the first pretrained language models for scientific text. Our follow-on work on domain adaptation via continued pretraining won an honorable mention for best paper at ACL 2020 🏆. I also develop methods for infusing visual layout information into models, culminating in PaperMage, an open-source Python library that won best paper at ACL 2023 System Demos 🏆.• I design rigorous evaluation guidelines for NLP, including few-shot learning and long-form text generation which won an outstanding paper award at EACL 2023 🏆.•.I’ve published some of the largest gold standard datasets for training and evaluating language models on scientific literature understanding tasks, including SciTLDR for summarization, SciFact for claim verification, Qasper for question answering, MultiCite for citation discourse understanding, and more.•.I’m a founding member and a lead contributor to the Semantic Reader project, which combines HCI and AI research to design intelligent reading interfaces for scientists. Through this project, I’ve developed novel systems including ScholarPhi which explains math notation, Scim which highlights salient passages, and PaperPlain which simplifies difficult passages and provides navigational guidance. One such system called CiteSee, which visually augments and personalizes inline citations while reading, won a best paper award at CHI 2023 🏆.

Seattle, Washington

• Improved network-wide MAPE by 30% using structural time series models with changepoint selection, Fourier decomposition of seasonal effects, and holiday effects. Used multiresolution data to augment forecasts for sparse streams.• Built pipeline for deploying and monitoring statistical models with automated retraining and prediction schedules. Currently used to compute time series forecasts on 1.5M+ daily data streams. Includes module for detecting outliers, changepoints, and gradual concept drift in model performance. Also includes module for simulating out-of-sample predictions.• Increased AUC 35% for streaming classifier that predicts buyers among 400M+ annual shoppers by adding active label querying and incorporating semi-supervised label propagation.• Prototyped tool for setting up, monitoring, and analyzing multi-website AB experiments using contextual mixed effects bandits and clustering using website similarity.• Placed 2nd at 2016 hackathon for SEO self-service tool that recommends ad keywords from among high-performing neighbors of a candidate word based on word embedding similarity.

Master Of Science (Ms), Statistics

• 2014-15 Dorothy M. Gilford teaching award

Bachelor Of Arts And Science (Bas), Statistics And Economics

• Graduated summa cum laude • Departmental citations in both Statistics and Economics