[Tech stack] Keras, GCP- Formerly Google X

[Tech stack] PyTorch, Remote Sensing APIs- Surveyed & implemented self-supervised methods to promote robustness to domain shift, boosted AUPRC by 32% on 10k images- Conducted pilot w/ SoTA CNN architectures for identifying processing plants in satellite imagery, 0.95 AUPRC on 2k images- Completed 2 ablation studies with variable data quality to estimate cost bounds for annotation & deployment, results critical for informing contract decisions

[Tech stack] Spark, Python- Reduced Siri's WER on Russian tail utterances by 6% using synthetic data injections to encourage morphology sensitivity in fixed-window neural LMs- Developed pipeline that ingests 3B+ lines of daily traffic to generate training data encoded w/ linguistic expertise

[Tech stack] Spark, Python, SQL, PHP- Created two-tower sparse NN embeddings that boosted user recommendation relevancy (now deployed, 6M+ hits per day)- Online experiments using learned embeddings reported a follow-through rate 13% higher than production version- Proposed a word2vec approach augmented with TF-IDF scoring that outperformed NN embeddings

[Tech stack] C++, Python- Wrote end-to-end code for a statistical learning algorithm (PhasED-Seq) for detecting circulating tumor DNA- Productionized PhasED-Seq's codebase for use as core method in Foresight Dx, a leading cancer relapse-detection startup, Series APublications:- Enhanced detection of minimal residual disease by targeted sequencing of phased variants in circulating tumor DNA (Nature Biotech 2021)- Integrating genomic features for non-invasive early lung cancer detection (Nature 2020)

[Tech stack] C++, Python- Increased runtime efficiency of 2 essential pipelines by 50x and 120x by implementing custom algorithms and data structures (ingests 100GB+ per run)

[Tech stack] C++, R, Docker- Designed, implemented and deployed a full software stack that went into production for 10,000 academic users- Built token-based authentication system for container distribution

[Tech stack] RoR, HTML/CSS, Javascript- Built an organization of 21 web & mobile developers - Developed a 3-month Ruby on Rails curriculum, aided 50+ students in becoming proficient in full-stack dev- Secured 15 projects (notable clients include school districts, healthcare clinics, and conservancy programs)

[Tech stack] RoR, OmniAuth, PostgreSQL- Deployed 3 web-based applications for clients using Ruby on Rails (6k+ active users in total)- Built client-side user interfaces with Javascript- Authenticated users & built relational databases