Computational Math in Industry and Beyond Seminar Series (CME 500) - Spring

Scott has been working with NASA’s Frontier Development Laboratory since 2017 to drive 18 separate projects that seek to use deep learning in the exploration of space and the preservation of Earth. In this fun talk, Scott describes why he cares so much about getting to Mars and how it led to his involvement with NASA. He starts with a tale of the rough first few weeks at NASA, confronted by skeptics. By carefully landing a key project, this nascent effort grew to multiple collaborations of NASA scientists, academics and Silicon Valley companies. These teams successfully applied deep learning to astrophysics, space weather, planetary science and more. Google of course was looking for a commercial outcome to pay for all this volunteer work. Scott goes on to describe how elaborate cocktails of AI -- brewed at NASA -- are now overtaking the mundane, driving 10x economic returns for customers, challenging the status quo of emerging, trillion dollar markets.

Scott is a member of Google Cloud’s CTO office (OCTO), a team of industry ex-CTOs who co-innovate with top customers and product teams. Scott reframes business processes as “tensors in, tensors out,” extracts data before and after a complex system, then builds and trains ML models to replicate, improve and optimize business processes. Scott currently focuses on the healthcare industry.

Scott has demonstrated efficacy in reverse engineering mainframe systems in healthcare claims processing, searching for exoplanets and minerals on the moon with NASA, and cutting the costs in millions of customer chat & voice interactions. Scott was blessed with the opportunity to build an amazing AI team within OCTO, where the team instigated call center AI, document AI, AI notebooks, and many of Google’s largest AI-first cloud deals. Scott holds a PhD in AI with multiple degrees from MIT and the University of Washington.

Previously, Scott landed public cloud at PwC for 200k employees in 2014, moved a video site for 5m users to AWS in 2008, sold a social photo site with 50M users in 2007, built mobile phone “widgets” in 2005, and launched a $13B web middleware and $4B web hosting business in the 90s. He’s an avid programmer, triathlete, space fan, guitarist, chef and father of two amazing daughters. Occasionally, Scott speaks in public. Follow @scottpenberthy on Twitter!

This lecture will talk about some of the key emerging industry trends in the world of AI and machine learning and their relevance to PayPal’s data science solutions which are deployed in areas such as Fraud and Credit risk assessment today. We will cover the trends in all areas of AI / machine learning – from the underlying AI/ML compute hardware and algorithms to the ecosystem of AI/ML tools and emerging need to build responsible AI systems. 

Vidyut joined PayPal in 2019 and leads the PayPal AI research team in the overall PayPal data science organization. His team’s charter consists of bringing in the latest that the world of AI / ML tech has to offer to PayPal. Before joining PayPal, Vidyut led AI research and engineering teams at NIO (a Chinese EV company) to build in house autonomous driving capabilities for NIO’s vehicles. Prior to that, he was at Qualcomm where he held several roles most recent being the Qualcomm sensor division AI / ML team lead. His team built and supported deployment of many sensor algorithms that run on today’s smartphones and smartwatches. He holds a PhD in Electrical Engineering from Cornell and is super passionate about everything AI!

Janice Tse is a Senior Director of Data Science at PayPal.  She studied Computer Science and has more than 15 years of work experience in the e-commerce and payment space. Janice is a strong believer in the power of data and Machine Learning to solve various technical challenges and drive better business decisions; she has taken on multiple roles ranging from customer behavior insights, marketing and fraud detection throughout her career.  Janice has also been actively involved in women in technology related initiatives to promote women presence in the industry.

This talk will be a flyover of projects that pair technical and linguistic background to tackle tricky problems in communication-focused products. 

Kathryn Hymes is a computational linguist and product leader with twin passions for math and language. She has spent a decade in technology leadership working on hard problems at the intersection of community and communication. Kathryn has built teams and products at Slack, Nextdoor and Facebook that pair computation and human language in novel ways across messaging, search, translation and localization. Outside the US, she has worked at Baidu HQ in Beijing, China, at CERN in Meyrin, Switzerland and as a Fulbright scholar in Budapest, Hungary. She did her graduate studies at Stanford in both ICME and Linguistics. Currently she is the head of International Product Expansion at Slack.

Geological sequestration of CO2 is one of the solutions to reduce atmospheric CO2 concentrations. Modeling and simulating CO2 injection in geological formations require advanced multiphysics, high-performance, numerical simulation tools. Lawrence Livermore National Laboratory, Stanford University, and Total have worked together since 2018 on creating such a tool. It is now available openly, and will be presented today.

Herve Gross is an R&D project lead in the Computing, Science, and Engineering program at Total Exploration & Production USA. He coordinates a partnership between Total, Stanford University, and Lawrence Livermore National Laboratory to develop a multiphysics reservoir simulator. This tool targets applications such as large-scale geological carbon storage simulation on exascale computing architectures. Herve holds an M.S. and a Ph.D. degree in Energy Resources Engineering from Stanford University.

In this talk, Danielle will discuss the importance niche of computational mathematics within industry, and discuss the core ICME skill-set in her work that is critical in devising novel and robust algorithms and implementing them within a production system from her work in time series forecasting at Amazon.

Danielle (Maddix) Robinson is a past ICME PhD alum, who is now a Senior Applied Scientist in Machine Learning Forecasting Group within AWS AI Labs.  In her PhD, she worked with Professor Margot Gerritsen on devising robust numerical methods to remove spurious temporal oscillations in the degenerate Porous Medium Equation.  She is passionate about the underlying numerical analysis, linear algebra and optimization methods behind numerical PDEs and applying these techniques to deep learning.  She joined AWS in 2018 shortly after graduating, and has been worked on statistical and deep learning models for time series forecasting, including combining physics-based approaches in the underlying ODEs and PDEs to model the dynamics with machine learning techniques.

Many of the products we use daily, and monetization of these products through advertising, are fully optimized through AI. State of the art AI is being born in the trenches of building these products, and not without substantial challenges. In this talk, we will explore how to build an AI product at scale. We will cover the spectrum from designing your customer success metrics, to using deep learning models at scale in production, to the necessity of unraveling bias in your data – all grounded in real-life examples of gotchas and highlights. We will conclude with a thought provoking look at causal modeling, and how it’s efficacy is being tested and strengthened in the context of real product problems.

Dr. Rukmini Iyer joined Microsoft in 2010 to advance conversational understanding for personal assistants and moved to Bing search advertising in 2012 to accelerate machine learning in the advertising product. Currently, she is CVP of the Advertising platform that serves many of Microsoft’s consumer facing products from search to news feeds. Her team manages key infrastructure and algorithmic components from advertiser automation and sales tools, to core NLP efforts for user, query and ad understanding, to machine learning of user and advertiser signals to run online ad auctions. Prior to Microsoft, Rukmini earned her PhD in Electrical Engineering from Boston University in 1998, working in speech recognition and NLP for the first 10 years at BBN Technologies and Nuance, and distributed ML and AI to improve online revenue and user experience at Yahoo. Her current interests include practical applications of reinforcement learning, improving long-term success of small businesses in Bing Ads ecosystem, and growing product teams that fundamentally embrace AI in design and development.

Analytics consulting is full of fun, challenging and impactful problems out in the real world, and relies on all the wonderful skills provided by ICME. Nicole will give a deep dive into a few projects she’s gotten to work on, including optimizing the network design of a large US bakery, scheduling production of bottles for a startup glass manufacturing company, and allocating spare parts across the data centers of a tech company.

Nicole Taheri graduated with her PhD from ICME in 2012 with a focus on optimization. After graduation, Nicole spent 4 years at IBM Research Ireland as an Optimization Research Scientist, working to optimize pump schedules for water utilities, create recommendations for ideal sales team structures and collaborate in the design of a semidefinite programming solver. 

Since 2016, Nicole has been at End-to-End Analytics (E2E), an analytics consulting company based in Palo Alto where she’s been able to implement optimization solutions at a number of large companies to help improve their business processes. E2E became a part of Accenture Applied Intelligence in March 2021, where Nicole continues to work on fun and challenging real-world analytics projects in a wide range of industry applications.

In this talk, Cris will review some of the highlights of his work at NVIDIA including low-communication FMM-accelerated FFTs, developing general tensor contractions for supercomputing, quantum computing, and deep learning, and how to use tensors to make programming the most complicated matrix multiplication algorithms easy.