Dec 2017 – Aug 2021 · 3 yrs 8 mos · San Francisco Bay Area

• Developed and patented AI algorithms for Linkedin Ads response prediction models, to compute optimal bid price for ad campaigns to increase LinkedIn’s revenue as well as improve advertiser’s ROI.
• Leading part of the team that conceived LinkedIn Audience Network (LAN), an ad platform for advertiser’s to display ads on third party apps/websites, in 2017. The product hit 1 Billion USD revenue in 2020.
• Designed and developed an automated platform to retrain and deploy ML models, reducing engineering effort from 2 weeks to 0 hours and eliminating manual errors.
• Demonstrated uplift in key business metrics such as revenue (2 Million USD per day), click through rate and conversion of ads by end to end implementation of new ML models.
• Implemented transfer learning from onsite (LinkedIn) ML models to offsite (third party) ML models to leverage a richer feature and improve prediction accuracy that boosted revenue by 5%.

Jul 2016 – Oct 2017 · 1 yr 3 mos · Austin, Texas Metropolitan Area

• Analyzed the performance of Machine Learning models like VGGNET16 to study the compute and memory requirements and evaluated the performance bottlenecks in hardware architecture using profiling tools.
• Ported Google’s open source software packages like Tensorflow, Bazel, Protobuf, gRPC to Oracle’s hardware by modifying source code to make it compatible on the new platform.
• Linked Oracle’s internal BLAS library to Tensorflow.
• Benchmarked and analyzed performance of Machine Learning frameworks like Sci-kit learn, Xgboost and 
Tensorflow on Oracle’s platform versus other platforms.
• Mathematically derived the theoretical flops/byte required for the Neural Network algorithm using stochastic gradient and batch normalization in order to design Machine Learning accelerators to speed up these algorithms.
• Applied Machine Learning principles to build tools to automate verification work done for Oracle’s hardware.
• Presented a technical talk on Machine Learning 101 for engineers working in other teams.

Feb 2016 – Feb 2017 · 1 yr · Berkeley, California

• Interview potential candidates for the Masters of Engineering program, UC Berkeley.