Presentations highlighted at the IT Summit 2024 AI Empowerment track

The 12th annual University of Arizona 2024 IT Summit took place over two days and provided attendees opportunities to network and explore the latest advancements in technology. Held on August 6th at the U of A Student Union Memorial Center, the summit also was livestreamed on August 7th. Throughout the summit, the Data Science Institute, CyVerse, ICDI, and other data science campus partners showcased their expertise during the AI Empowerment track. Attendees gained valuable insights into AI, technology, and data science innovations. If you are interested in learning more about the presentation or the topic, contact the presenter. 

Tuesday August 6th Seeing Beyond the AI Hype Cycle Collaborate in the Cloud with CyVerse So You Want to Build an AI Application? AI Working Group Reports

Wednesday, August 7 Day in the Life of a Data Scientist Building your own Chatbots with a Purpose UArizona & NSF Infrastructure for AI Making Cloud Infrastructure Accessible with CACAO Sensor Fabric: AI from Wearable Data Do More with your Data Using RAG: Building Chatbots Teaching infrastructure for AI/ML Learning & Maker Space for ML/AI: UArizona Data Lab Campus AI Practitioners Community Building

Aug. 6th details for select sessions on the AI Empowerment track

10:30 - 10:50 AM Seeing Beyond the AI Hype Cycle

• Nirav Merchant, Director, Data Science Institute

Looking to keep up with the dizzying pace of AI innovations when they're often closed source and proprietary? Fortunately, there is a vibrant ecosystem of open source tools and platforms with permissive licenses, coupled with no cost and highly scalable AI infrastructure available through NSF ACCESS and the recent National AI Research Resource (NAIRR) initiative. With these NSF resources alongside campus computational  infrastructure, IT and AI expertise, U of A researchers and students can build AI powered solutions that meet the  Open Science and reproducibility requirements from federal agencies.  We all have an important role to play in shaping AI innovation as creators, consumers and educators. This requires community engagement, hands-on training for pragmatic and informed adoption of AI capabilities as they mature over time.

---

1:30 - 1:50 PM Collaborate in the Cloud with CyVerse

Get an overview and demonstration for how CyVerse Data Store and cloud automation platforms are democratizing access to scalable commercial and private cloud resources. The CyVerse Data Store integrates with any S3 compatible storage service—like AWS Simple Storage Service and the Open Storage Network (OSN)—to allow users to share their data securely with collaborators at any institution while keeping their S3 credentials private. Access is easy through web and desktop tools, and container (docker) based analysis tools.  If you are launching cloud-based virtual machines in services like AWS Elastic Compute Cloud (EC2) or NSF’s CloudBank and Jetstream2, or NIH STRIDES, save time and make your work reproducible by automating deployment. CyVerse Cloud Automation and Continuous Analysis Orchestration (CACAO) combines the Infrastructure as Code (IaC) capabilities of Terraform and Ansible to create recipes that are readily deployed through an easy to use and simple web interface for end users.

---

2:00 - 2:20 PM  So You Want to Build an AI Application?

• Nirav Merchant, Director, Data Science Institute
• Shravan Aras, Assistant Director for Sensor Analysis & Smart Platforms, Center for Biomedical Informatics & Biostatistics 

Learn where to start! AI based applications typically require a significant amount of computational and data resources, and higher rigor in software engineering and data management. Find out how to make effective use of campus and NSF resources, alongside platforms like Hugging Face, to explore AI models and build and deploy solutions that span campus and commercial resources.

---

3:00 - 3:45 PM  AI Working Group Reports

Aug. 7th details for select sessions on the AI Empowerment track

9:00 - 9:50 AM  Day in the Life of a Data Scientist

---

10:00 - 10:50 AM  Building your own Chatbots with a Purpose

Large Language Models (LLM) and Generative Pre-trained Transformers (GPT) with chat interfaces like ChatGPT have ushered a new era for how users are interacting with complex knowledge platforms. There are many tangible benefits in integrating these LLM and GPT based solutions with data from sources that are not readily accessible or available on the public internet. Multiple groups at the University of Arizona are prototyping solutions where they have successfully developed applications that combine the power of these technologies. Get highlights from some of the use cases and demonstrations, followed by a panel discussion.

---

11:00 - 11:50 AM  UArizona & NSF Infrastructure for AI

Discover the physical and virtual resources available for AI teaching, training, and research from the Research Computing group in UITS and from University of Arizona's owned and operated CyVerse, an NSF project. Learn what current resources are freely available to all University of Arizona staff, faculty, and students and how to gain access to them. Find out also about national AI efforts, including the National Artificial Intelligence Research Resource Pilot (NAIRR) and Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS).

---

1:00 - 1:20 PM  Sensor Fabric: AI from Wearable Data

• Shravan Aras, Assistant Director for Sensor Analysis & Smart Platforms, Center for Biomedical Informatics & Biostatistics

Wearables are an essential part of multiple research projects and studies, but researchers struggle in managing data from sensors and deriving value from it. Learn about Sensor Fabric and its architectural design for making data AI ready. Find out how the presenters developed the Sensor Fabric to address many of the common challenges faced for ingesting wearable sensor data from Fitbit, Apple Watch, Sensor Kit, Empatica, etc. into a scalable and cost-effective cloud-based solution. This includes use of object storage such as S3 and schema catalogs such as AWS Athena. See how data from Sensor Fabric can be efficiently and quickly used to create AI models written in PyTorch and also using AutoML techniques offered by both GCP and AWS, along with connections to open-source visualization tools running in CyVerse.

---

1:00 - 2:50 PM  Making Cloud Infrastructure Accessible with CACAO

Learn how to leverage CACAO (Cloud Automation and Continuous Analysis Orchestration), an open-source, cloud-native platform, to use infrastructure-as-code (IaC) or domain-specific language (DSL) templates for your deployments. Simplify the challenges with deploying complex software installations and configurations on multiple virtual machines or cloud services, including those with specialized hardware like GPUs. Make manual installation less time-consuming and error prone. With templates, your deployments are also reproducible. Hear about best practices from multiple NSF projects that are utilizing CACAO for a range of activities, including cloud access for science and engineering, ML-based simulations, data science workshops in Ecology and Geospatial analysis, and LLM based workflows and inference.

---

1:00 - 2:50 PM  Do More with your Data Using RAG: Building Chatbots

Get an introduction to Retrieval-Augmented Generation (RAG) systems within our campus environment, revolutionizing how we manage and utilize information in the realm of generative AI. Explore their transformative potential in streamlining workflows, facilitating knowledge retrieval, and fostering innovation. See how they work in successful deployments, such as at KMap, and find out the strategies employed and lessons learned in the process.  Examine RAG system architecture, including the mechanics of vector databases, semantic search capabilities, and effective prompting methodologies.

---

1:30 - 1:50 PM  Teaching infrastructure for AI/ML

• Tyson Swetnam, Co-PI and Science Lead, CyVerse
• Chris Reidy, Manager, Research & Discovery Technologies, UITS

Do you need to include hands-on AI/ML learning material in courses and workshops? ML/AI based assignments need specialized software stacks (Jupyter, MLFlow, Pytorch etc.) with dependencies that make them hard to manage and maintain. Additionally, they often require specialized hardware like GPUs and significant data handling capacity.  Find out about the resources available to faculty wanting to develop new creative courses and assignments. Learn common practices that ease the burden of managing the ML/AI software resources, and how to utilize no cost infrastructure available through NSF ACCESS and UA HPC.

---

2:00 - 2:50 PM  Learning & Maker Space for ML/AI: UArizona Data Lab

• Carlos Lizarraga, Computation and Data Science Educator, Data Science Institute
• Brenda Huppenthal, GRA, Computer Science, UArizona Data Lab
• Megh Krishnaswamy, GRA, Linguistics, UArizona Data Lab
• Chan “CK” Chi-Kwan, Associate Astronomer, Steward Observatory

Do you need to stay on top of the evolving landscape of machine learning (ML) and artificial intelligence (AI)? Learn and experiment with new developments and advancements at the UArizona DataLab. Participate in codefests/hackathons on cross-cutting topics of interest. This hub and clearinghouse for the U of A community provides:
• extracurricular and non-formal training through short format workshops in ML/AI tools and technologies
• consultations on research projects for incorporating ML/AI methods into research lab workflows
• assistance in onboarding and training students and staff who are involved in implementing ML/AI solutions
Gain an understanding of the progress and roadmap for non-formal learners to utilize DataLab, and for departments and other units to participate in DataLab to showcase their activities and highlight the vibrancy of ML/AI learning opportunities offered across campus.

---

3:00 - 3:50 PM  Campus AI Practitioners Community Building

• Michael Dorland, IT Support Analyst, Campus IT Partnerships
• Barney Maccabe, Executive Director, Institute for Computation and Data-Enabled Insight
• Ash Black, Director AI & Industry, Institute for Computation and Data-Enabled Insight/AI Core
• Shravan Aras, Assistant Director for Sensor Analysis & Smart Platforms, Center for Biomedical Informatics & Biostatistics
• Darcy van Patten, Chief Technology Officer
• Chan “CK” Chi-Kwan, Associate Astronomer, Steward Observatory

Hear from the panel on steps taken to bring together the University community using AI. Be part of the conversation for next steps.