Saurabh Agrawal Minimalism, Degrowth, Wellness

Staff Data Scientist

Agoda

Feb 2023 - Jun 2023 (5 months)

• Hypothesis Testing on A/B experimentation data - Identified that one of the key assumptions the team had been using to build ML models didn't seem to hold true. • Finding duplicate entities - Done as part of the hiring assessment. Task was to match entity descriptions across two sets of descriptions that potentially represented the same underlying entity (hotel) despite variations in description due to noise or alternate representations in title, address, city name etc. Around 400 examples of matching entity descriptions were provided. Proposed AI solution trained an SVM classifier and could find entity matches with >98% accuracy and ~100% recall. • Relevant Tech Stack: Python, Numpy, Scipy, Scikit-Learn, Pandas, Matplotlib

Principal Data Scientist

Walmart

Aug 2017 - Sep 2021 (4 years 2 months)

• Recommender Systems - Led the design and development of product recommendation AI using machine learning on customer and product data. Tech Stack was Python/Numpy/Scipy/Scikit-Learn/XGBoost/Pandas/Matplotlib. Systems partially running on GCP. • Responsibilities included identifying data sources, creating datasets amenable to ML, formulating the problem as a supervised or unsupervised learning problem, feature extraction and instance labeling, model selection, training and validation, choosing appropriate evaluation metrics, A/B experimentation, technical presentations in various internal forums, leading and mentoring a team, debugging issues in live model performance doing RCA.

Staff Engineer

Walmart

Jan 2013 - Jul 2017 (4 years 7 months)

• Worked on significant parts of Affiliate facing Walmart Open API. Designed and implemented a pipeline for ingesting catalog updates into a catalog database near real-time. Proposed and implemented an ElasticSearch based backend for catalog paginated API which could be queried for parts of catalog using various filters on product attributes. The proposed solution was a big improvement over what it replaced: a monolithic zip feed of the entire catalog. • Anomaly detection - Proposed a solution for identifying clickbait in PPC (pay per click) affiliates. The solution once implemented immediately found that two of the largest affiliates (by commission volume) were sending clickbait traffic. • Display Retargeting Backend - Display retargeti

Co-founder

Startup

Jan 2011 - May 2012 (1 year 5 months)

Built a movies recommendation engine where a user selected at least 3 titles they liked to get an infinite scrolling recommendation feed. Implemented content filtering algorithms, built autocorrect and search functionality using Apache Lucene inverted indexing which allowed searching for movies by various attributes.

Software Engineer

Playdom

Mar 2010 - Jun 2010 (4 months)

Implemented a few small features in the game backend using php and mysql.

Software Engineer

Yahoo!

Jul 2008 - Feb 2010 (1 year 8 months)

Java enterprise development in the EWS (Enterprise Web Services) project

Saurabh Agrawal

About

Detail

Timeline

Résumé