• Developed novel algorithms to automatically generate labels to train machine learning models for DNA sequencing which resulted in an average 0.3% increase of sequencing accuracy over previous labeling methods

• Researched and developed methods to classify sensors which helped identify optimal biochemical parameters for the company’s next-generation sequencing platform

• Created filtering algorithms to remove poor quality sensors to prevent them from being basecalled, resulting in about an 8% reduction in storage requirements and computing time in the analysis pipeline

• Provided data analyses to perform system characterization, hypothesis testing and to help design targeted experiments to root cause anomaly observations and to validate hypotheses.

• Increased productivity and drove decision making by developing statistical pipelines that provided insights into the error modes of the company’s DNA sequencing methods

• Built NextClimb, an interactive web-app (in Python and Flask) hosted on AWS for climbers that recommends climbing routes based on implicit and explicit user preference data scraped from MountainProject.com

• Combined collaborative filtering based on matrix factorization and content-based filtering using natural language processing to create a hybrid recommender

• Reconstructed hold-out user-ratings to within 0.6 on a user rating scale of 0-4 (over 13,000 users) and provided relevant recommendations (up to 87% mean-average precision, choosing among over 25,000 climbing routes)

• Met tight deadlines by finishing this project from start to finish (idea conception, data gathering/cleaning, model selection/validation, and deployment) within 3 weeks