Advise the Flywheel Founders and engineering team on machine learning theory and applications.

Ranking and deep learning models related to questions, answers, and topics.

• Created and helped ship a machine learning algorithm aimed at dynamically creating search queries from publisher web pages, increasing search relevancy by 2X for some web domains.• Designed and conducted experiments with the online crowd-sourcing platform CrowdFlower to measure search result relevancy, employing various metrics such as nDCG and Kendall rank correlation.• Built a relevancy dashboard that showcased mobile-advertisement relevancy, involving a) a daily PySpark job, and b)… Show more • Created and helped ship a machine learning algorithm aimed at dynamically creating search queries from publisher web pages, increasing search relevancy by 2X for some web domains.• Designed and conducted experiments with the online crowd-sourcing platform CrowdFlower to measure search result relevancy, employing various metrics such as nDCG and Kendall rank correlation.• Built a relevancy dashboard that showcased mobile-advertisement relevancy, involving a) a daily PySpark job, and b) a front-end developed using tornado, JavaScript / JQuery, AJAX and Bootstrap.• Planned and analyzed results of an AB test focused on click-through rate, employing t-tests andcorrection for multiple-comparisons.• Used an internal wikipedia data index and other existing tools to design and prototype a short-textnamed entity recognition and disambiguation system.• Prototyped a machine learning model that predicted whether a disambiguated named entity was”relevant” to a publisher web page. Show less

-Partnered with San-Francisco-based startup company URX to build a custom Named Entity -Recognition (NER) system using a Conditional Random Fields classifier that outperforms the established Stanford NLP standard by 50% for miscellaneous entities-Created a web app (http://entitizethis.info) that demos the NER classifier, and demonstrates the effectiveness of the NER system for document search using Elastic Search.

-Using graph theory and network analysis to derive novel insights about community organization of brain structures, using fMRI data. -Using transcranial magnetic stimulation to study the role of prefrontal regions in our ability to recognize previously encountered stimuli see: http://despolab.berkeley.edu/bbowles2/.-built a web app to visualize brain lesions in patients with traumatic brain injury and stroke.

Currently using high-resolution fMRI of medial temporal lobe subregions in combination with multivariate statistical techniques to derive novel insights about underlying mechanisms of memory encoding and retrieval.

Applied signal-detection models to better understand how people recognize famous names; supervised 3 senior honors psychology research projects; gave guest lectures in advanced statistical concepts; collaborated with clinicians to design memory tests for patients

Used general linear models and multivariate pattern classification to analyze brain imaging data and infer group differences between clinical and neurotypical populations

Received a research award to conduct a 4-month survey study in Cambodia aimed at better understanding transportation barriers to healthcare in over 150 families; conducted focus groups, interviews, and contextual inquiries, and filed a completed report to the Cambodian government

Wrote Python scripts to complete automated web-searches in established online databases to identify proteins with specific amino acid sequences; information pertaining to each protein was organized and reported by consulting additional online resources