Hi all, and welcome to my home on the web! I’m Luis, a Software Leader at Waymo, making cars understand agent behavior by scaling large multimodal models, using state-of-the-art deep learning algorithms, and deploying large-scale, distributed infrastructure.

I was previously at Meta AI (formerly known as Facebook), re-engineering Meta AI’s deep learning platforms for vision, text, and audio. Before that, I worked on graph neural networks at DeepMind, advancing the state-of-the-art in traffic prediction and learned the art of large-scale distributed systems at Google by building Procella. I’ve also given several lectures at North Carolina A&T University and developed the curriculum for Algorithm Foundations for AI and Cybersecurity

I spend my free time doing research for fun.

I have a MS in Computer Science from Stanford, advised by Doug James and working with Jurij Leskovek. I specialized in artificial intelligence and CS theory - focusing on convolutional neural networks and sequence models for image and language understanding and graph neural networks for geometric deep learning. I also have a BA from Harvard University in computer science and mathematics.

I am the co-author of The Effectiveness of Data Augmentation in Image Classification using Deep Learning and ETA Predictions with Graph Neural Networks in Google Maps. I worked to take my research to production, substantially improving the travel-time predictions in Google Maps (featured across several media outlets such as CNBC, Engaget, VentureBeat, CNET, the Verge, and ZDNet).

Latest Work/News

• Checkout my work on the Metaverse at Meta AI driving our 0->1 efforts for Horizon Worlds to help creators build on our next computing platform.
• Algorithm Foundations for AI and Cyber-security is now officially public, with all content and material freely available! Thanks to everyone at Stanford for helping out!
• torchrecipes has been open-sourced! Check-out the github.!
• Our work on re-engineering Meta AI’s deep learning platforms to be more interoperable.
• Our work on improving travel time predictions for Google Maps was accepted to CIKM and is now out!

My work and mission

I use qualitative, quantitative, and computational methods to holistically investigate and solve large-scale problems across products used by billions. Most of my work focuses on building technical solutions to technical problems. I love working to promote positive user content by teaching machines to understand text, image, and video. I like making positive social impact, such as reducing carbon emissions from vehicles by developing novel routing algorithms for Google Maps, and improving YouTube’s video recommendations through advanced deep learning techniques.

I live by the following principles:

• Focus on Long-Term Individual Growth: Maximize potential for growth (people skills, technical skills, leadership skills) to build strong/clear foundation to “change the world” (eg, being a lever that makes the difference)
• Create Positive External Impact: My work must make people’s lives better, provide a clear positive service and add direct value. It is not enough to work on something “big” or “important” - the work has to be a net positive for the world.
• Build Great Systems: I work on teams that are leaders - focused on building great products for users, great systems for us. My teams are not “fast followers” or “catching-up” - we are leading because we are creating and building, working on hard problems, and pushing the boundaries of our domain. This also applies to non-technical support systems for individuals to grow.
• Be Kind: Be direct, be empathetic, be supportive. How people feel matters, how the team feels matters. I work to build a culture of support and collaboration - avoiding individual competition and instead inspiring each other to be better.
• Do The Right Thing: We do what’s right, even if it’s more costly or difficult. We correct and monitor progress, but our guiding principle is to do the right thing.