Suggestions
Alexander Morrise, PhD
Head of Graph Data Science and AI, building Louie.ai
Alexander Morrise, PhD, is the Head of Graph Data Science at Graphistry, Inc., where he leads efforts in machine learning, focusing on AI, natural language processing, and graph neural networks.12 He is currently spearheading the development of Louie.ai, a generative AI data platform that simplifies data interaction and analysis through natural language, aiming to democratize complex data analytics.13
Dr. Morrise has a diverse background spanning academia and industry. He was previously an Assistant Professor of Theoretical Physics, specializing in Quantum Gravity and String Theory, before transitioning to AI applications in industry.12 His academic journey includes a PhD in Theoretical Physics from UC Santa Cruz, where he studied topics in theoretical physics and mathematics, and carried out research in String Theory.2
In his role at Graphistry, Dr. Morrise oversees the development of Graphistry[AI], a toolkit that transforms various data sources into graph AI visualizations and models for sectors like cybersecurity and supply chain optimization.1 He is also involved in projects such as winning the U.S. Cyber Command AI Alert Data Challenge using graph-based approaches.15
Prior to joining Graphistry, Dr. Morrise held several positions in the tech industry, including:
- Chief Technology Officer & Cofounder at STAY OPEN HOTELS
- Lead Data Scientist at Beats Music
- Founder and CEO of TensorML
He has also served as an advisor for various companies, including The Consilience Project, Arternal CRM, MediaJel, and 8tracks.2
Dr. Morrise's expertise lies in machine learning, data science, and their applications in complex data analysis. His work at Graphistry and with Louie.ai focuses on making data analysis more accessible and powerful through the use of generative AI and advanced visualization techniques.34
Highlights
Sol Eun is #hiring. Know anyone who might be interested? Sol is one of the best.
In the world of deep learning, the concepts of equivariance and invariance often present a formidable challenge. The crux of the matter lies in the difficulty of finding or approximating the equivariant transform. However, insights from Category Theory, which illuminate the nuances of box-(non)-commutativity, might just provide the clarity we need.
But why does this matter in the practical realm? Let's consider a few scenarios:
- Cost Functions: When designing cost functions for your models, do they generalize across various products or services? Or do you find yourself needing a unique function for each distinct 'class'? The ability to create a universal cost function could significantly streamline your processes.
- Portfolio Yield Prediction: Is there a singular function capable of predicting the yield on your portfolio given market fluctuations? The quest for such a function is akin to the search for an equivariant transform in deep learning.
- Model Building: In the intricate dance of model building, the presence of 'super selection' sectors can be the difference between success and failure. These sectors can lead to months of misguided ML engineering, leaving you puzzled about where things went awry.
Understanding and leveraging these concepts can elevate your approach from merely navigating through the complexities of deep learning to strategically harnessing its potential. In a world driven by data and algorithms, mastering equivariance and invariance could be your key to unlocking new horizons.
