• Contributed in the Acceleration Enablement team of 30+ members to develop software stacks for training Pytorch ML Models on custom hardware ASICs.

• Created new kernels and LLVM compiler bindings in open source PyTorch codebase.

• Transformed the ads revenue model pipeline deterministic nature.

• Reduced the 10% difference in output with same input to no difference.

• Saved on average 9 employee hours per week per team. (C++, SQL)

• Implemented features in Trading platform Atlas and increased 2% internal trading to save the stamp duty fees in the stock market for clients (Java).

• Developed the framework for more robust and automated testing and increased testing coverage (Java, Groovy)