Thomas Kidd Email & Phone Number

San Francisco, CA, US

San Francisco, CA, US

Sunnyvale, California, US

I work on the Weight Streaming compiler stack at Cerebras Systems, where I design and implement solutions to enable the efficient training of very large (100B+ parameter) machine learning models on Cerebras hardware.

Sunnyvale, California, US

I work on the Weight Streaming compiler stack at Cerebras Systems, where I design and implement solutions to enable the efficient training of very large (100B+ parameter) machine learning models on Cerebras hardware.

Sunnyvale, California, US

Currently working on "brain-scale" neural network architecture at Cerebras: https://spectrum.ieee.org/cerebras-ai-computers

Sunnyvale, California, US

During my internship at Cerebras I:- created a numeric analysis tool which algorithmically identifies all sources of numeric errors between TensorFlow and Cerebras kernels (i.e. the tool can identify the exact Op where numeric divergence begins between the original TensorFlow model and the compiled, low-level Cerebras model), saving days to weeks of.

Waterloo, Ontario, CA

Undergradute Research Assistant with University of Waterloo's Configurable Architectures Group (WatCAG). Developing cutting-edge Machine Learning Accelerators and accompanying software frameworks for FPGAs.

Santa Clara, California, US

C++ SWE and HWE on the Deep Learning Acceleration team at Intel CSWEDeveloped scalable, high-throughput IP in SystemVerilog, built and tested accompanying hardware models in C++, and debugged ML-graph compiler issues.

Santa Clara, California, US

HWE Intern on the FPGA Accelerated Libraries team at Intel PSGCreated novel, high performance linear algebra kernels in OpenCL and Intel HLS. Researched and developed prototype sparse linear algebra hardware design patterns. Developed C++ models to estimate resource usage and optimze design parameters to maximize throughput.

Santa Clara, California, US

HWE Intern on the FPGA Accelerated Libraries team at Intel PSGAuthored Matrix-Inverse and Triangular Matrix-Multiply hardware kernels in OpenCL and Intel HLS. Developed C++ plugins to verify functional correctness and numerical accuracy of the hardware kernels. Overhauled regression testing framework, resulting in a significant savings of engineering time.

Jakarta, DKI Jakarta, ID

iOS software developer for the Contacts team at BBMAuthored low-latency features to enable real-time "New Friend Joined" notifications. Assisted in design pattern refactorign to significantly reduce technical debt.

Toronto, ON, CA

Provides automation of infrastructure services and data analysis to ensure the accuracy of various data sets. Additional roles include working with third party vendors to develop custom solutions and automation processes.