As a member of the ChatGPT team, I managed the massive fleet of GPUs serving ChatGPT. I enabled rapid iteration on our language models while maintaining the performance and reliability of the ChatGPT service. Ownership of ChatGPT's GPU fleet: I was responsible for managing the substantial fleet of GPUs that perform inference in order to serve ChatGPT. I maintained the health of the overall fleet, enabled the testing and rollout of new language models, and supported major feature launches. Tooling for GPU fleets at scale: I led an initiative to automate and streamline management of our GPU fleet at such unprecedented scale. I proposed and implemented tools that reduced the number of manual operations, resulting in faster iteration on model testing/deployment. I collaborated with other teams to align their roadmaps with ChatGPT's tooling needs. A/B testing framework for models: I played a pivotal role in establishing our A/B testing framework for evaluating new models, which was crucial for decision-making on model rollouts and for identifying model regressions. I contributed to the selection of key performance metrics, the implementation of metrics and traffic splitting, and the analysis of A/B experiment results.

I'm a tech lead manager on the Log Applications team, which builds services that enable Argo teams to manage logs from self-driving vehicles and run workflows on these logs in the cloud. Each month our logs grow by ~1 PB, resulting in ~1 million new labeled events that are accessed through our services. I built and own a service that enables Argo to replay past vehicle event logs using more recent versions of the self-driving software in a scalable manner. This service is now used to process on average over 100,000 logged events per day, and it is critical for supporting ML training and for validating the safety of our cars. The team I lead owns the service that acts as Argo's repository for storing, editing, and searching testing data, including simulation scenarios and tests based on vehicle logs. As the team lead I work with 38 other teams across the autonomy, offboard infrastructure, and testing organizations. I set the vision and priorities for the team. I am responsible for providing high-level designs and design reviews for both my team and others. As manager I manage my team's projects, and over the last 8 quarters we've successfully delivered on 96% of our quarterly goals (which are designed to be ambitious stretch goals). I support my reports and foster their career growth through 1-on-1's and performance reviews.

Joined as the 5th employee and led backend development, including scaling from zero customers to being installed on over a billion devices in under 4 years. Built the data pipeline to provide customers with analytics for hundreds of millions of daily active users. Promoted to lead architect, overseeing all design decisions, including designing the end-to-end interactions between the backend, SDKs, and frontend. Spearheaded machine learning initiative: defining the roadmap, preparing user data, and training/evaluating models on Google Cloud. Led backend engineering team and fostered career growth of team members, conducting performance reviews and 1-on-1's.

Geo Search Quality Evaluation: Developed tools for evaluating search quality on Google Maps and Google.com. These tools played a critical role in aiding development of algorithm changes, supporting launch decisions, and directing future efforts. Local Structured Data and Structured Search: Worked on structured data for local businesses, with projects ranging from infrastructure for modeling and serving structured data, to introducing new and better ways of using this data in search.

As a member of the ChatGPT team, I managed the massive fleet of GPUs serving ChatGPT. I enabled rapid iteration on our language models while maintaining the performance and reliability of the ChatGPT service. Ownership of ChatGPT's GPU fleet: I was responsible for managing the substantial fleet of GPUs that perform inference in order to serve ChatGPT. I maintained the health of the overall fleet, enabled the testing and rollout of new language models, and supported major feature launches. Tooling for GPU fleets at scale: I led an initiative to automate and streamline management of our GPU fleet at such unprecedented scale. I proposed and implemented tools that reduced the number of manual operations, resulting in faster iteration on model testing/deployment. I collaborated with other teams to align their roadmaps with ChatGPT's tooling needs. A/B testing framework for models: I played a pivotal role in establishing our A/B testing framework for evaluating new models, which was crucial for decision-making on model rollouts and for identifying model regressions. I contributed to the selection of key performance metrics, the implementation of metrics and traffic splitting, and the analysis of A/B experiment results.

I'm a tech lead manager on the Log Applications team, which builds services that enable Argo teams to manage logs from self-driving vehicles and run workflows on these logs in the cloud. Each month our logs grow by ~1 PB, resulting in ~1 million new labeled events that are accessed through our services. I built and own a service that enables Argo to replay past vehicle event logs using more recent versions of the self-driving software in a scalable manner. This service is now used to process on average over 100,000 logged events per day, and it is critical for supporting ML training and for validating the safety of our cars. The team I lead owns the service that acts as Argo's repository for storing, editing, and searching testing data, including simulation scenarios and tests based on vehicle logs. As the team lead I work with 38 other teams across the autonomy, offboard infrastructure, and testing organizations. I set the vision and priorities for the team. I am responsible for providing high-level designs and design reviews for both my team and others. As manager I manage my team's projects, and over the last 8 quarters we've successfully delivered on 96% of our quarterly goals (which are designed to be ambitious stretch goals). I support my reports and foster their career growth through 1-on-1's and performance reviews.

Joined as the 5th employee and led backend development, including scaling from zero customers to being installed on over a billion devices in under 4 years. Built the data pipeline to provide customers with analytics for hundreds of millions of daily active users. Promoted to lead architect, overseeing all design decisions, including designing the end-to-end interactions between the backend, SDKs, and frontend. Spearheaded machine learning initiative: defining the roadmap, preparing user data, and training/evaluating models on Google Cloud. Led backend engineering team and fostered career growth of team members, conducting performance reviews and 1-on-1's.

Geo Search Quality Evaluation: Developed tools for evaluating search quality on Google Maps and Google.com. These tools played a critical role in aiding development of algorithm changes, supporting launch decisions, and directing future efforts. Local Structured Data and Structured Search: Worked on structured data for local businesses, with projects ranging from infrastructure for modeling and serving structured data, to introducing new and better ways of using this data in search.