Aug 2022 – May 2025 · 2 yrs 9 mos · Sunnyvale, California, United States · Hybrid

Sep 2020 – Aug 2022 · 1 yr 11 mos

• I am a research assistant in computer science department at USC. I am actively looking for data science opportunities.
• Website : nsl.usc.edu/aastha

May 2020 – Aug 2020 · 3 mos · New York, United States

• Deep Learning models for temporal sequences.
• Used industrial Scale data to train to train complex pipelines
• Imbalanced data was used to train LSTM family classifiers

Jan 2018 – Jul 2018 · 6 mos · Singapore

• Supported all the functions (Marketing, A/B Testing, Data warehouse design, Dashboard Suite) for the upcoming Grab Logistics business.
• Tools : SQL, Tableau, Python
• End to End responsibility of analytics of Grab-Logistics business
• Launched and analyzed the A/B test for multiple simultaneous bookings feature from Grab Logistics across Indonesia, Vietnam and Philippines
• Created dashboard suite including dashboards for daily business health metrics, customer NPS, customer retention reports, hourly and area wise - book through rate analysis, elementary customer abuse and fraud detection
• Deep dived into allocation metrics and recommended improvements to improve the first round allocated rides
• Trained the new members of the team to use Tableau and SQL

Jun 2016 – Jan 2018 · 1 yr 7 mos · Bangalore Urban, Karnataka, India

• Worked for the Trust and Safety Team to build analytical rule based models to detect and prevent customer fraud in Ratings and Reviews. Significant contribution to reducing customer *return* fraud on the platform.
• Tools : SQL, Hadoop, QlikView
• Rating and Review Fraud detection and prevention
• Defined and sized the universe of rating and review fraud
• Defined velocity limits to auto-reject spam ratings and reviews on the platform
• Implemented the “Share Act” algorithm in R; Conceptualized a graph based tool to detect fraudulent communities in the user graph
• Reduction of Customer Return Fraud
• Defined thresholds to classify customer returns into fraud and abuse buckets.
• Defined granular auto-reject thresholds for high abuse and fraudulent customer returns; Rules based on order risk and customer risk scoring
• Estimated cost savings from the rules - $150 K per month
• Open Box Delivery Pilot
• Created customer profile segment for targeted open box delivery to decrease missing and mis-shipment return of high value items
• Simulated improvement in the return rates; Pilot run successful
• Miscellaneous
• Hackathon Finalist, used customer comments to classify returns requests (team of 4)

Jul 2014 – Jul 2016 · 2 yrs · Bangalore Urban, Karnataka, India

• Part of the Digital Marketing and Analytics team
• Analyst supporting six different Private label card partners (website launch, rewards program, operations dashboard)
• drive to digital interstitial campaign analysis and A/B test launches
• created BRDs (Business Requirement Docs) for feature launches, analysed and created insights from customer navigation journeys on website, customer dropout funnels on forms, verified javascript tags, database design for storing them, analyzed interstitial campaigns, SEO campaigns
• Worked on several cost cutting business cases including drive to paperless, cross-LOB data sharing analysis,
• Net Present Value (NPV) estimation for calculating marketing campaigns spends