In this episode, hosts Mariana Menchero and Faranak Golestaneh explore the cutting-edge world of foundation models for time series forecasting with guests Azul Garza Ramírez, cofounder of Nixtla, and Mononito Goswami, one of the developers of MOMENT, a family of open-source foundation models for general-purpose time series analysis. The conversation delves into the backgrounds of these innovators and their journey into the realm of time series analysis and forecasting.

The podcast explores the guests' transition into working with foundation models for time series forecasting. The guests describe the empirical approach they took, inspired by the success of Transformers in other domains like video, images, and text. Their experiments with adapting these models to time series data yielded exciting results, leading to the development of new products and tools.

The conversation sets the stage for a deep dive into the challenges and opportunities presented by foundation models in time series forecasting. The discussion highlights the need for massive, diverse datasets and the potential for these models to learn patterns and extrapolate to new data effectively.

This episode underscores the rapid advancements in time series forecasting and the growing importance of foundation models in pushing the boundaries of what's possible in this field. It offers listeners a glimpse into the minds of innovators who are shaping the future of time series analysis and its applications across various industries.

In this episode, hosts Mariana Menchero and Faranak Golestaneh explore the cutting-edge world of foundation models for time series forecasting with guests Azul Garza Ramírez, cofounder of Nixtla, and Mononito Goswami, one of the developers of MOMENT, a family of open-source foundation models for general-purpose time series analysis.

In this episode, we discuss the guests' transition into working with foundation models for time series forecasting. The guests describe the empirical approach they took, inspired by the success of Transformers in other domains like video, images, and text. Their experiments with adapting these models to time series data yielded exciting results, leading to the development of new products and tools.

The conversation sets the stage for a deep dive into the challenges and opportunities presented by foundation models in time series forecasting. The discussion highlights the need for massive, diverse datasets and the potential for these models to learn patterns and extrapolate to new data effectively.

This episode underscores the rapid advancements in time series forecasting and the growing importance of foundation models in pushing the boundaries of what's possible in this field. It offers listeners a glimpse into the minds of innovators who are shaping the future of time series analysis and its applications across various industries.

In this episode, guest hosts George Boretos and Arian Sultan Khan explore the intersection of Neuroscience with Forecasting & AI with guest Kai Markus Mueller, acclaimed neuroscientist and a pioneer in Neuropricing.

Kai, who began his journey in psychology with aspirations of becoming a child psychotherapist, eventually shifted his focus to cognitive psychology and neuroscience. His transition from academia to the industry led to the invention of Neuropricing that utilizes fMRI and EEG to understand consumer behavior and predict responses to advertising and pricing.

The podcast delves into Kai’s innovative work, highlighting how brain activity can often predict consumer behavior more accurately than traditional self-reported methods, with success stories such as Starbucks coffee pricing research and Pepsi’s strategy in Turkey.

Kai explains the practical applications of neuroscience in business, such as storyboard testing for advertising effectiveness. He discusses the integration of AI with neuroscience to enhance predictive models.

He also shares insights on balancing his various roles as an entrepreneur, professor, and industry practitioner, emphasizing the importance of a supportive team. Looking ahead, Kai sees immense potential for neuroscience and AI to transform business strategies, pricing, and drive marketing success.

The conversation underscores the growing mainstream acceptance and practical benefits of these advanced technologies.

In this episode, we had the privilege of hosting Mitchell O'Hara-Wild, data scientist and lead developer of the widely used and highly acclaimed forecasting packages, Fable and Feasts.

Mitchell is a PhD candidate at Monash University, Australia. He shared insights on a wide range of topics, including his journey into data science and forecasting, the reasons behind the development of the popular Fable package, and his views on AI in forecasting.

We also discussed Mitchell’s research on DAGs (Directed Acyclic Graphs) in the context of forecast reconciliation, as well as his consulting experience forecasting COVID-19 cases in Australia. Moreover, we had the opportunity to talk about his experience delivering workshops to researchers and practitioners through the IIF's Forecasting for Social Good community (F4SG) and at useR! conferences.

Listen to this podcast and learn more about Mitchell’s remarkable work in the realm of forecasting, software development, and the future of forecasting in the era of AI.

In this episode, we spoke to Joannes Vermorel, founder and CEO at Lokad, a quantitative supply chain software company.

Joannes discussed how supply chain theory is broken down, and that we need to think in terms of paradigms and modules rather than models for solving supply chain problems. He talked about issues in time series forecasting and judgmental forecasting. He emphasized how critical it is to have a holistic view of the problem, to aim for optimization of the entire system. and to acknowledge that we often don’t know the metric to be optimized and it requires some experimentation. We also discussed how Lokad is deploying AI pilots to address some of the important problems in supply chain.

To learn more about Lokad, visit https://www.lokad.com/ or check them out on YouTube.

In this episode, we spoke to Laurent Ferrara, Professor of International Economics at SKEMA Business School. Laurent discussed the role of nowcasting, particularly in the realm of macroeconomic nowcasting. He delved into the details of the models and methods that have been proven effective in this domain. Laurent also talked about GDP nowcasting using Google data and shared some intriguing results from his recent research.

Laurent is the program chair of the 44th International Symposium on Forecasting, which will be held in Dijon, France. He provided an overview of the conference program and explained why we should attend!

In this episode of our podcast, we delve into the intricate world of machine learning (ML) deployment with Dr. Eric Siegel, author of the book AI Playbook, Mastering the Rare Art of Machine Learning Deployment.

Dr. Siegel, once an avid advocate of ML, now approaches the field with a disciplined yet optimistic perspective. He shares invaluable insights on how businesses can effectively implement ML strategies. Our discussion revolves around a range of compelling topics, from the inspiring story of Jack from UPS, who leveraged his psychology background to revolutionize parcel delivery, to the common pitfalls that cause many ML projects to fail.

Eric elucidates the six crucial steps for ML deployment, emphasizing the importance of ethical considerations in this rapidly evolving field. Whether you're a student, a business leader, or just an AI enthusiast, this episode offers a treasure trove of knowledge and strategies to navigate the complex landscape of machine learning deployment.

Our guests are Michele Trovero, leader of the Forecasting R&D group at SAS, and Spiros Potamitis, Data Scientist and Product Marketing Manager at SAS. We delved into the intriguing intersection of Language Model-based AI (LLMs) and forecasting software.

We explored the openness of forecasting software providers to embrace LLMs and discussed the profound impact these models could have on the industry. Michele and Spiros shared insightful examples of LLM applications. They elaborated on the way code generation capabilities powered by LLMs would enhance the development of forecasting software and the user experience. Additionally, they explored how LLMs could democratize forecasting, and discussed other tools and technologies that could contribute to this goal. We also discussed the typology of models behind LLMs, and their applicability in forecasting, as well as the limitations and enablers in using AI-pretrained models in forecasting. The discussion extended to SAS Visual Forecasting and Model Studio, shedding light on their functionalities and workings.

Michele and Spiros speculated on the areas of focus for forecasting software companies, enhanced automation in forecasting, shifts in user consumption patterns, and anticipated integrations between forecasting systems and other technologies.

They recommended the following for further study:

1. How Will Generative AI Influence Forecasting Software? by Michele Trovero and Spiros Potamitis, Foresight: The International Journal of Applied Forecasting.
2. A Glimpse into the Future of Forecasting Software, by Spiros Potamitis, Michele Trovero, Joe Katz, Foresight: The International Journal of Applied Forecasting.