Training robots to understand the world around them using modern AI. Building an amazing team.(Company was formerly known as Positronix.)

I joined Amazon with the personal mission of bringing deep learning to Amazon. In 2013 this was a crazy idea, and while I can hardly take credit for the obvious eventual success, I personally accomplished a number of important and highly profitable things along the way.1. Applied Science: I kicked off Amazon's first deep learning projects for product similarity, visual search, and recommendations systems. For each of these projects I convinced relevant leadership to take a bet on deep learning, provided guidance on the science, and helped with the engineering.2. Amazon SageMaker: I designed, prototyped, and helped deliver the Docker-based architecture for distributed deep learning training in Amazon SageMaker. This was based on challenges I saw in applied science, and as the engineering lead for AWS’s first ML service, Amazon Machine Learning.3. GPU hardware: I organized all of Amazon’s deep learning scientists to successfully lobby for the creation of more powerful GPU instances leading to the P2 instances in Amazon EC2.4. AutoML: I built the Automatic Model Tuning feature in Amazon SageMaker that uses ML to optimize ML models. I defined the initial product, led the engineering effort, and guided the science team.5. Apache MXNet: I brought MXNet to Amazon, by first identifying it as intrinsically well-suited for Amazon's needs, helping with key hires, and championing its internal adoption.6. Open source: I took Amazon's deep learning framework DSSTNE to the public. Scott Le Grande wrote all the code (I helped with requirements). But I made the case to leadership to let it out the door, under the Apache 2 license. This was the first time Amazon released significant original software under the permissive Apache 2 license, and thus represents a small milestone of a cultural shift.7. Reinforcement Learning: I led the launch of SageMaker RL in 2018, the first cloud managed reinforcement learning solution, and was a key tech lead on the DeepRacer project.

BNCO is a small 501c3 that supports orphaned children in Ethiopia through direct sponsorships and by operating a community medical clinic. For more info see http://bluenile.org/

Built technology to support a large team of social media analysts.

Launched Google's video chat service. Drove integration of instant messaging with mobile.

Led a cross-functional team to rebuild Rhapsody's win32 client as an online music streaming site, which back then was still a pretty novel thing.

Back in the very early days of the web, we built sites that were very advanced for the time like e-commerce systems and maps you could scroll in your browser.