Divyansh Agarwal Email & Phone Number

New York, Ny, Us

Graduate school

Palo Alto, Ca, Us

Worked on improving the personalization of search ranking. Dove deep into Glean's proprietary pagerank algorithm to improve personalization and move topline search metrics.Also did the following to help with data science efforts within the company:- Created new metrics to help better analyze searching ranking performance and experiments- Improved data foundations for analytics efforts and search ranking evaluation- Analytics to support growth and customer success initiatives.Languages: Java, Golang, Python, SQL

San Francisco, California, Us

I worked on two teams during my tenure - Shared Rides Matching, and Product Ranking.I worked on UberPool and UberX Share while I was on the Shared Rides Matching team.My biggest achievement in this role:I helped build UberX Share from scratch, and played an important role from the Applied Science in shaping this product from day 1 - for a period of 1.5 years, I was the only IC Applied Scientist working on this product.I also ran and analyzed the first experiment that measured the value of Shared Rides (Uber Pool and UberX Share) to Uber's business. This work helped the Shared Rides team make the case for its continued need at Uber post-COVID.

Berkeley, Ca, Us

Worked with Prof Bin Yu and Prof Avi Feller, under the mentorship of PhD students William James Murdoch and Elijahu Ben-MichaelDeveloped Novel Approaches for Causal Inference using Deep Learning, and consequently reduced bias on Kang-Schafer simulation to 1/6th of the bias associated with Entropy Balancing, a widely used Causal Inference approach. Worked on Interpreting Neural Networks used for NLP Tasks such as Question-Answering by adapting Contextual Decomposition (developed in-house for interpreting neural networks) to these specialized Neural Networks. Subsequently analyzed the patterns learned by high-performing Deep Learning approaches on the SQuAD dataset.

Berkeley, Ca, Us

Various roles in course staff for Data Science Courses (Data 8, Data 140) and Stats Courses (Stat 135)

Berkeley, Ca, Us

Modeled student behavior and cognition on MOOCs from EdX and Khan Academy using LSTMs and Deep Knowledge Tracing respectively.Supervised directly by Prof Zach Pardos

Mountain View, Ca, Us

Data Science in AdvertisingUsed Causal Inference approaches to optimize Ads volume on Quora’s Ads platform, while accounting for the trade off between user engagement and Ads revenue.Also built a heuristics based classifier to detect India-specific questions during a company hackathonTools Used: Python (Pandas, Numpy, Seaborn), SQL, Redshift, Airflow

Redwood City, California, Us

- Identified users based on how they pick up their smartphone with near 0% Equal Error Rate using Dynamic Time Warping- Improved performance of gait cycle detection algorithm by 65%, by implementing smart heuristics- Classified 150 daily human activities from sensor data using Convolutional Neural Networks with 76% accuracy

Mumbai, Maharashtra, In

- Analyzed customer complaints against financial institutions in the USA - Predicted number of complaints against a bank in each state using Linear Regression with a R-squared value of 0.93- Predicted whether response to a customer complaint will be timely using Logistic Regression with AUC of 0.75

Bachelor'S Degree, Computer Science And Statistics

Master Of Science - Ms, Computer Science

Education record