Jan 2023 – Present · 3 yrs 3 mos

Sep 2022 – Nov 2024 · 2 yrs 2 mos · Mountain View, California, United States

• Led AI Frameworks (JAX, PyTorch, TF, Keras, Gemini Frameworks), AI Foundations (XLA Compiler, Pathways, PJRT, Triton, Mosaic) and OnDevice ML (AppliedML, TFLite, MediaPipe) teams powering AI tech, LLMs across Google, DeepMind, GCP and Android Platforms.

Jan 2019 – Sep 2022 · 3 yrs 8 mos

• Led PyTorch Core, the frameworks and platform powering the machine learning across the world driving the transformation of PyTorch driving AI research to production innovating across model authoring, model exploration, efficiency and scaling building production ecosystems.

Jan 2017 – Jan 2019 · 2 yrs · San Francisco Bay Area

• • Led all Core Platforms in SF ranging for Onboard Infrastructure, Sensor Software, Data Infrastructure, Performance and MLPlatform for the AV stack. My teams drove strategy and execution of the end-2-end AV software systems stack.

Jan 2011 – Jan 2017 · 6 yrs · San Francisco Bay Area

• Managed and led a team responsible for prototyping, delivering and productizing new runtime compiler software stack for Metal 1.1 on Mac OSX.
• Key leader in the collaboration efforts with Apple's strategic hardware partner to launch the new Apple Graphics Metal API on Mac OSX at 2015 WWDC.
• Worked closely with external partners like Adobe to port their apps to Metal leading up to 8x performance and power improvements.

Jan 2009 – Jan 2011 · 2 yrs · San Francisco Bay Area

• Led cross-company teams focused on end-to-end development including prototyping and delivering OpenCL support on Intel GPU Mac platforms. Collaborated with consumer and pro application teams to use OpenCL to improve performance and user experience.
• Established and implemented future product roadmap and direction for GPGPU software stack for Apple platforms. Built and led a team to prototype software technology for GPGPU/Heterogeneous Compute projects working with both hardware and software apps teams.
• Drove performance optimization for OpenGL workloads via compiler and software optimizations improving the performance of application and games in Mac platforms.

Jan 2006 – Jan 2009 · 3 yrs · Austin, Texas Area

• Founding member of the AMD GNU Tools (gcc, binutils and glibc) team responsible for optimizing performance on current and future x86 and AMD processors.
• Led a team involving 4 full time engineers implementing support and optimization for AMD Opteron (Barcelona and Orochi) processors in gcc, binutils and glibc.
• Key contributor for evaluating and publishing software optimization techniques tuned towards AMD processors.
• Technical Liaison for Redhat: Empower collaboration between the two GNU tools teams. Responsible for setting up face to face meetings and giving presentations and tutorials on current and next generation AMD processor architecture improvements

Jan 2004 – Jan 2006 · 2 yrs · Austin, Texas Area

• One of the founding members of the open source compiler team at AMD.
• Worked on analyzing applications and benchmarks thereby improving performance of compilers and future AMD processors.
• Developed fast math/string library routines that achieve maximum performance on AMD processors.
• Implemented a brand new infrastructure in C++ for testing both performance and functionality of all the AMDCoreMathLibrary routines. The test infrastructure included unit tests, benchmarks and applications which run in both single thread and multiple threads.
• Improved the performance of AMD’s internal cycle accurate simulator by about 23% with the help of both source code and compiler optimizations. This resulted in increasing the efficiency of several design and performance teams within AMD.

Jan 2003 – Jan 2003 · 0 mo · Greater Minneapolis-St. Paul Area

• Worked on developing NLP extraction techniques (in Java Annotation Patterns Engine) to extract information from large number (in millions) of documents.