Daniel Jerónimo Gómez Antonio Email & Phone Number

Seattle, Wa, Us

Seattle, Wa, Us

Seattle, Washington, United States

- Lead an initiative to speed up initialization of distributed training using MP with +1024GPUs. Collect and analyze profiles to identify the bottleneck. Research, scope, define anddevelop a solution, which included a distributed store (key-value pair) based on RDMA.Results showed improvement of 6x from 12 to 2 minutes, they were presented tostakeholders, including a discussion with a Sr. PE to drive upcoming projects.- Design, implement and optimize an algorithm for a collective communication primitive (allreduce) within GPUs for distributed training. Delivered a modular architecture that wasadopted by the team to power other collectives. Performance boost of 10% against baselineand 90% test coverage. Technologies: C++, Python (PyTorch), CUDA, RDMA.- Develop a feature to improve resiliency during distributed training. Design and implementa service to identify GPU failures during training and orchestrate an automatic recoveryprocess. The service operated at different layers, from data plane to control plane usingC++, Go and Java. As a result, training could recover from last saved checkpointautomatically.

Seattle, Washington, United States

- Work in design and implementation of engines for AI to improve distributed data parallel training and make it faster and cheaper.

Austin, Texas, United States

- Own the design and development of a .NET application to monitor and analyze network stack processes.- Mentor a software intern during summer. Coordinate and define requirements for the project, which was successfully completed. Guide and provide continuous feedback.

Austin, Texas

- Work as software developer to provide driver support for hardware products intended for automotive Ethernet testing.- Design, implement and maintain driver software, firmware and and low level utilities using different technologies like C++, C, Python and LabVIEW.- Develop middleware to interact with the driver and add support for multiple network related protocols.- Automate complex testing involving multiple devices and high speed transfers to reduce test time of product benchmarks from days to hours. Also, integrate results into a web dashboard to monitor data over time.- Serve as part time SCRUM master to facilitate work items to the team, keep the backlog organized to reflect the team commitments and expectations. Organize and lead SCRUM ceremonies including stand-ups, plannings and retrospectives.

Austin, Texas Area

I was a software engineer intern, I worked in the drivers team. My project was related to implement a communication protocol to an existing driver that interfaces NI hardware with LabVIEW. I learnt driver developing in user and kernel mode using C++.

Santiago De Querétaro, Querétaro, Mexico

I was a software engineer intern at Digitum, mostly doing full-stack tasks focused to web development.On the daily basis, I used HTML, PHP, Javascript, AJAX, MySQL and Ruby.The main project I worked on was related to automate an application for the marketing department.

San Pedro Garza García, Nuevo León

I worked doing basic web development and databases task for the IT department.