Avesh Singh Email & Phone Number

San Francisco, Ca, Us

Feature Store

San Francisco, California, Us

Cardiogram uses machine learning to detect signs of heart disease from wearables like the Apple Watch and Garmin. Interested in learning more? We're hiring!I lead the machine learning team in training models to predict atrial fibrillation, diabetes, hypertension, and sleep apnea. Publications:DeepHeart: Semi-Supervised Sequence Learning for Cardiovascular Risk Prediction (AAAI 2018)Passive detection of atrial fibrillation using a commercially available smartwatch (JAMA Cardiology, May 2018)Highlights:-Setup data pipeline (5TB of postgres data -> 20GB TensorFlow input) and Google Cloud training infrastructure. -Collected 350k labels from userbase. -Maintained rapidly growing Heroku Postgres instance (1.5 TB/month growth). -Mentored intern in building and releasing Cardiogram for Garmin.

Mountain View, Ca, Us

Software engineer (L4) on Search Quality --we figure out what order the search results show up in.Highlights:• Lead engineer in generating signals to rank Knowledge Engine results alongside web results, impacting 13% of traffic-weighted queries.- Built pipeline to obtain signals for all Knowledge Engine results from 1.2 petabytes of raw data- Worked with Web Answers and Local teams to generalize these signals for use in ranking of their results (collaborated with 3 teams, held workshops, worked with PMs, wrote documentation)• Instructor for Life of a Query Noogler orientation class.- Taught 100 new Googlers how the search stack works.• Developed and implemented algorithm to detect contextual optionalization. For example, in the query [mozzarella cheese recipe], cheese is optional as it is implied by mozzarella.- Processed 70 terabytes of data to come up optionality scores for optionality(foo | {bar, baz, quz}).- Utilized this novel signal to increase precision of semantic query clustering (eg "how tall is barack obama" = "obama height" != "barack obama age).- Experimented with reweighting query terms• Owner of the Komodo Web Corpus to store query-keyed data (impacts top-5 results for 16% of queries)- Took over orphaned code that hadn't been touched in 1.5 years.- Improved runtime of the bottleneck job from 3+ days to 1 hour.- Guided two teams in adding to the corpus.• Mentored 3 interns, 1 full time engineer.

Mountain View, Ca, Us

Built a machine learning model to improve accuracy of policy violation detection in Google AdWords. Explored methods to improve trademark violation detection.

Generated analytics on customer energy usage for monthly Energy Reports. Created and deployed daemon to schedule and monitor data importing. Operationalized analytics databases for robustness and security.

Pittsburgh, Pa, Us

TA for Principles of Functional ProgrammingTaught weekly recitation to 25 students to reinforce and further explore lecture content. Created and graded lab assignments. Hosted review sessions and assisted students during office hours.

Master'S Degree, Computer Science

Bachelor Of Science, Computer Science

Education record