DSI

AI Deep Dive: AI-Powered Curricular Intelligence: Transforming Medical Education Through Precision Learning

Miguel Nunez — Mon, 02 Mar 2026 01:57:49 +0000

On February 27th, we hosted an AI Deep Dive Session in collaboration with the ��ͷ�� University School of Medicine (VUSM) and the Department of Biomedical Informatics with from ��ͷ�� University Medical Center, where we explored how retrieval-augmented generation (RAG) and large language models can be grounded in institutional medical curricula to deliver personalized, adaptive learning at scale. Dr. Stenner, Associate Dean for Education Design and Informatics at VUSM and an AMA ChangeMedEd Innovation Grant recipient, led a discussion on building a multimodal AI platform that transforms how medical students learn, how faculty teach, and how institutions govern medical education.

Highlights:

Purpose: The project aims to address the integration crisis in preclinical education, where content is siloed by discipline while clinical practice demands cross-domain reasoning. By grounding LLMs in ��ͷ��’s complete first-year curriculum, including 150GB of lectures, slides, handouts, and transcribed audio, the platform enables contextual, authoritative AI support that generic tools cannot provide.
Focus Areas: The platform organizes 20 features into five strategic clusters: a Precision Learning Engine for adaptive study support, a Clinical Thinking Accelerator to build physician reasoning from day one, a Social Learning Ecosystem that uses AI to facilitate peer learning and normalize struggle, a Curriculum Intelligence Platform for automated accreditation mapping, and a Faculty Empowerment Suite to reduce administrative burden while improving teaching quality.
AI Applications: Core technical approaches include multimodal RAG pipelines for ingesting diverse curricular content, knowledge decay detection with automated spaced repetition, clinical reasoning scaffolding through structured frameworks, concept relationship mapping via interactive knowledge graphs, and automated curriculum alignment auditing using semantic analysis.

Session Insights:

The group explored critical implementation challenges around optimal chunking and embedding strategies for multimodal content, as well as metadata schemas that need to serve both learner-facing citation and institutional curriculum mapping simultaneously.
Discussion centered on privacy-preserving learner modeling, recognizing that tracking individual knowledge decay and engagement patterns raises important questions about data governance, consent, and the responsible use of educational analytics.
The session surfaced key questions about sustainable cost modeling for LLM inference at full class scale and evaluation frameworks that can satisfy both internal proof-of-concept milestones and external grant applications, with the two-year implementation timeline targeting a nationally disseminable model for precision medical education.

Conclusion:

The AI Deep Dive with Dr. Shane Stenner and the ��ͷ�� University School of Medicine showcased how AI can move beyond generic chatbot functionality to become deeply integrated curricular intelligence, transforming fragmented medical education into personalized, adaptive learning experiences. This session provided a unique opportunity for those interested in medical education, biomedical informatics, and AI-powered learning systems to engage in meaningful discussion and collaboration.

Are you interested in hosting a future AI Deep Dive? Contact us at datascience@vanderbilt.edu.

The post AI Deep Dive: AI-Powered Curricular Intelligence: Transforming Medical Education Through Precision Learning appeared first on DSI.

AI Deep Dive: AI-Powered Self-Care: Exploring Voice AI and Mobile Solutions for Rural Heart Failure Patients

Miguel Nunez — Mon, 09 Feb 2026 21:47:49 +0000

On February 6th, we hosted an AI Deep Dive Session in collaboration with the ��ͷ�� School of Nursing with (PhD, RN, FAAN), Associate Professor of Nursing in the Health Promotion, Populations, and Health Systems Community. The ��ͷ�� School of Nursing is one of the nation’s premier graduate nursing schools, recognized for excellence in education, practice, and research with a strong commitment to serving underserved populations. Dr. Stolldorf, a Fellow of the American Academy of Nursing, specializes in implementation science and the sustainability of health care innovations, with a focus on improving patient safety and quality of care. She led a discussion on leveraging AI technologies to expand the GUIDED-HF telehealth self-care intervention to rural heart failure patients who face significant barriers to traditional telehealth.

Highlights:

Purpose: The GUIDED-HF intervention has demonstrated effectiveness in supporting heart failure self-care through telehealth, but rural patients face unique barriers—limited internet connectivity, unfamiliarity with video conferencing, and lack of technical support at home. The session explored AI-powered alternatives to rethink how the intervention is delivered to these underserved populations.
Focus Areas: The discussion centered on whether Audio OpenAI (voice AI) could enable patients to engage with the intervention through natural phone conversation, and how smartphone apps and chatbot solutions compare in terms of accessibility, patient engagement, and intervention fidelity for users with phone-only internet access.
AI Applications: The session examined voice AI for conversational intervention delivery, chatbot-based patient interaction, and mobile app solutions for daily symptom tracking—all designed to work within the constraints identified through rural patient interviews and surveys.

Session Insights:

Insights from rural patient interviews shaped the technical requirements: patients primarily rely on phones for internet access, struggle with telehealth setup without in-person assistance, and want simple tools for daily health monitoring. Any AI-powered solution must balance accessibility with intervention fidelity.
The collaboration between DSI and the School of Nursing enabled a multidisciplinary brainstorming session on technical architecture, user experience design, and practical implementation considerations—combining nursing expertise with data science and AI capabilities.
The session identified key considerations for future grant submissions, including infrastructure requirements and which delivery modality best serves low-tech literacy users while maintaining the structured nature of the heart failure self-care intervention.

Conclusion:

The AI Deep Dive with Dr. Stolldorf and the ��ͷ�� School of Nursing showcased how AI-powered voice, chatbot, and mobile technologies can help bridge the digital divide in rural healthcare delivery. This session provided a unique opportunity for those interested in voice AI, mobile health applications, and rural healthcare to engage in meaningful discussion and collaboration.

Are you interested in hosting a future AI Deep Dive? Contact us at datascience@vanderbilt.edu.

The post AI Deep Dive: AI-Powered Self-Care: Exploring Voice AI and Mobile Solutions for Rural Heart Failure Patients appeared first on DSI.

AI Deep Dive: Automating Training Analytics for Elite Soccer Performance

Miguel Nunez — Tue, 27 Jan 2026 19:07:44 +0000

On January 23rd, we hosted an AI Deep Dive Session in collaboration with ��ͷ�� Athletics with , Head Coach of ��ͷ�� Women’s Soccer, where we explored how AI and computer vision could transform the collection and analysis of training data for elite athletes. Coach Ambrose has built ��ͷ�� into one of the premier programs in the SEC since arriving in 2015—the 2018 SEC Coach of the Year has guided the Commodores to seven NCAA Tournament appearances, including the program’s first-ever No. 1 seed and Elite Eight appearance in 2025. His teams have won two SEC Tournament championships (2020, 2025), the 2018 SEC regular season title, and produced seven All-Americans while maintaining exceptional academic standards with eight Scholar All-America honorees. The program’s data-driven culture has demonstrably improved player performance, raising team shot-on-target percentage from 38% to 51% through targeted training feedback.

Highlights:

Purpose: The program seeks to automate labor-intensive manual video tagging of practice sessions, which currently requires student analysts to watch every practice recording and tag individual events for each player—limiting the frequency and depth of feedback coaches can provide.
Focus Areas: While commercial platforms serve game analytics well, the critical gap lies in training data—the daily practice sessions where player development actually happens. The ultimate vision is a pipeline that generates individual player dashboards within hours of training completion rather than days.
AI Applications: Exploring whether modern multimodal AI models can identify soccer events and attribute them to individual players from practice video, with key technical challenges including player identification without jersey numbers and integration with the existing Spideo camera system.

Session Insights:

The session explored which soccer events are easiest versus hardest for AI to detect, with shots and goals likely more accessible than tackles, 1v1 duels, and pass completions—informing a phased implementation approach.
Hosting the session at the McGugin Center allowed participants to see the program’s facilities firsthand and understand how an automated system would connect with Spideo’s camera infrastructure and existing dashboard tools.
Discussion covered whether to build a labeled dataset of human-tagged practice video to fine-tune a specialized model versus relying on prompt engineering with general-purpose models.

Conclusion:

The AI Deep Dive with Darren Ambrose and ��ͷ�� Women’s Soccer showcased a compelling opportunity for AI to address a real operational challenge in elite athletics—transforming manual video review into automated, same-day performance insights. This session provided a unique opportunity for those interested in sports analytics, computer vision, and applied AI to engage in meaningful discussion about giving ��ͷ�� a sustained competitive advantage in athlete development.

Are you interested in hosting a future AI Deep Dive? Contact us at datascience@vanderbilt.edu.

The post AI Deep Dive: Automating Training Analytics for Elite Soccer Performance appeared first on DSI.

AI Deep Dive: Mapping Early Adopters with AI: Predicting B2B Beachheads for ��ͷ�� Startups

Miguel Nunez — Tue, 21 Oct 2025 18:53:16 +0000

On October 17th, we hosted an AI Deep Dive Session in collaboration with the Owen Graduate School of Management’s Center for Entrepreneurship with , Director of the Center for Entrepreneurship at ��ͷ��. The Center for Entrepreneurship (C4E) supports ��ͷ�� founders at every stage through structured programs, funding, and mentorship. Webb, a seasoned entrepreneur who founded MEDarchon (acquired by XSOLIS) and holds multiple patents in healthcare technology, led a discussion on using AI to help B2B startups identify their first customers.

Highlights:

Purpose: Address the critical challenge of identifying “beachhead” customers—the early adopters essential for new ventures to survive and cross the chasm to broader market adoption.
Focus Areas: Building an end-to-end system that operationalizes go-to-market discovery, with particular attention to edtech and healthcare sectors where buyers are consolidated.
AI Applications: Integrating EDGAR Form D signals for pre-seed/seed companies, automated web scraping to track customer logos over time, and firmographic data enrichment to train machine learning models that forecast look-alike prospects.

Session Insights:

The session explored best practices for ethically and reliably collecting web and third-party data, including considerations around scraping frequency and data quality maintenance.
Collaboration between data science and entrepreneurship faculty can turn noisy market signals into rigorous, founder-friendly insight that ��ͷ�� B2B founders and campus programs can use to prioritize outreach.
Discussion addressed evaluation frameworks that tie predictions to real outcomes, helping founders validate whether AI-generated prospect lists translate into actual customer acquisition.

Conclusion:

The AI Deep Dive with Baxter Webb and the Center for Entrepreneurship showcased how AI can transform the guesswork of early-adopter identification into a repeatable, data-driven process. This session provided a unique opportunity for those interested in data science, entrepreneurship, and venture development to engage in meaningful discussion and collaboration.

Are you interested in hosting a future AI Deep Dive? Contact us at datascience@vanderbilt.edu.

The post AI Deep Dive: Mapping Early Adopters with AI: Predicting B2B Beachheads for ��ͷ�� Startups appeared first on DSI.

AI Deep Dive: Democratizing Survey Analysis with AI-Driven Interactive Tools

Miguel Nunez — Tue, 27 May 2025 18:48:23 +0000

On May 23rd, we hosted an AI Deep Dive Session in collaboration with the Department of Political Science and the ��ͷ�� Poll with from the College of Arts and Science, where we explored how AI can transform the way scholars, journalists, policymakers, and the public interact with survey data. Clinton, the Abby and Jon Winkelried Chair and Co-Director of the ��ͷ�� Poll who also serves as Senior Election Analyst for NBC News, outlined a vision for “Pollscape”—a web app and large-language-model interface designed to make ten years of ��ͷ�� Poll results accessible through intuitive, conversational experiences.

Highlights:

Purpose: Democratize survey analysis by creating a plug-and-play platform that empowers non-coders to explore, visualize, and discover insights from poll data without specialized statistical training.
Focus Areas: Building transparent, conversational interfaces that can ingest questionnaires, toplines, and micro-data while surfacing related historical questions and subgroup breakdowns.
AI Applications: Leveraging large language models to auto-generate appropriate visualizations, suggest connections across historical surveys, and enforce statistical guardrails to prevent misleading inferences.

Session Insights:

The discussion explored AI pipelines for cleaning, harmonizing, and indexing ten years of bi-annual state and annual Nashville survey data into a coherent, searchable system.
Collaboration between political science and data science expertise is essential for building platforms that balance user accessibility with statistical rigor.
Pollscape aims to scale beyond ��ͷ�� Poll data—ultimately allowing any researcher to upload a survey and have the platform handle the heavy lifting of analysis and presentation.

Conclusion:

The AI Deep Dive with Josh Clinton showcased how AI can revolutionize survey transparency and public access to polling data. This session provided a unique opportunity for those interested in political science, data science, HCI, and public opinion research to engage in meaningful discussion about making rigorous analysis accessible to all.

Are you interested in hosting a future AI Deep Dive? Contact us at datascience@vanderbilt.edu.

The post AI Deep Dive: Democratizing Survey Analysis with AI-Driven Interactive Tools appeared first on DSI.

AI Deep Dive: AI-Driven Pet Nutrition Innovation

Miguel Nunez — Tue, 06 May 2025 18:36:02 +0000

On May 2nd, we hosted an AI Deep Dive Session in partnership with Mars Petcare with from Mars Pet Nutrition, where we explored how artificial intelligence is transforming product design and consumer insights in the pet food industry. Mars Petcare is a global leader in pet nutrition with brands including Pedigree, Royal Canin, and Whiskas, and has invested over $1 billion in digital innovation and AI-powered tools to enhance pet health outcomes. Chin-Ping Su, Senior Manager of Open Innovation at Mars Pet Nutrition, led a discussion on developing intelligent systems that diagnose performance issues, formulate recipes, and predict pet parent responses.

Highlights:

Purpose: Explore how AI can enhance pet nutrition innovation by building intelligent systems that leverage extensive technical data frameworks.
Focus Areas: AI-driven product design, performance diagnostics, and consumer response prediction in pet food development.
AI Applications: Development of AI agents capable of diagnosing product performance issues, generating improvement solutions, assisting in recipe formulation, and anticipating pet parent perceptions.

Session Insights:

The session highlighted how comprehensive technical data frameworks can power AI agents that streamline the product development cycle from formulation to market validation.
Mars Petcare’s collaboration with DSI provided a unique opportunity to discuss how industry-scale AI applications are reshaping pet nutrition science and consumer engagement.
Attendees explored the future potential of intelligent systems that bridge the gap between nutritional science and pet parent expectations.

Conclusion:

The AI Deep Dive with Chin-Ping Su and Mars Petcare showcased the transformative potential of AI in pet nutrition, from enhancing product performance to understanding consumer needs. This session provided a unique opportunity for those interested in AI applications, consumer products, and data science to engage in meaningful discussion and collaboration.

Are you interested in hosting a future AI Deep Dive? Contact us at datascience@vanderbilt.edu.

The post AI Deep Dive: AI-Driven Pet Nutrition Innovation appeared first on DSI.

ChatGPT-O3 Reasoning Agents Unlock Long-Horizon Multimodal Problem Solving

Miguel Nunez — Wed, 30 Apr 2025 20:37:55 +0000

AI Flash: ChatGPT-O3 Reasoning Agents Unlock Long-Horizon Multimodal Problem Solving

Event Overview

The latest AI Flash session at ��ͷ��’s Data Science Institute—hosted by Chief Data Scientist Jesse Spencer-Smith—pulled back the curtain on ChatGPT-O3, OpenAI’s newest “reasoning model.”

Unlike earlier releases that respond the moment a prompt arrives, O3 thinks first—planning a chain of reasoning, then selectively calling tools (Python, web search, image processing, automations, memory, and more) before it speaks. That extra deliberation, paired with 200 billion parameters, a 200 k-token context window, and native multimodality, lets O3 tackle complex problems that once took researchers weeks.

Breakthrough Capabilities

Long-Horizon Reasoning: O3 can stay on task for 10–20 minutes (or more) without “losing the thread,” continuously updating its plan as new evidence arrives.
Autonomous Tool Use: When text alone isn’t enough, the model writes and runs its own Python, browses the web, crops and enhances images, or stores interim notes in memory—then reasons over the results.
Native Multimodality: Text, images, and (in future) audio are tokenized together, so the model “looks” at pixels while it “reads” words—no fragile hand-offs between separate vision and language systems.
Steerability & Transparency: Users can reveal the model’s private chain-of-thought, correct wrong assumptions on the fly, and explicitly direct which tools to employ.

Live Demonstrations

“Where Was This Toad?” – O3 deduced that a mysterious backyard photo was shot in Puerto Rico by identifying a cane toad, consulting the user’s travel history, and cross-checking regional species maps—solving a puzzle the user couldn’t crack unaided.
Campus Photo Forensics – Given a group selfie in front of ��ͷ�� residence halls, the model zoom-cropped laptop stickers, adjusted contrast, and compared brickwork patterns before concluding the shot was on Alumni Lawn.
Eye-Blink Research Pipeline – In 30 minutes O3 drafted, coded, and benchmarked multiple computer-vision strategies (edge detection, adaptive thresholding, CNN segmentation) to extract eyelid-motion metrics from terabytes of IR footage—work a Ph.D. team estimated would take a month.
Measuring Belief-System Distance – For a project in formal epistemology, the agent produced a landscape of Euclidean and non-Euclidean metrics, suggested Finsler geometry for asymmetric belief revision, and generated a reading list—all in one pass.
Historical Tech-Policy Sleuthing – It uncovered overlooked declassified sources on Robert McNamara’s Vietnam “electronic barrier,” then drafted FOIA request templates that cite exact box numbers to accelerate National Archives retrievals.

Why It Matters

O3 blurs the line between assistant and collaborator. By reasoning with images, code, and external knowledge—then iterating for minutes, not milliseconds—it can:

Short-circuit weeks of literature review, data wrangling, or prototype coding.
Act as a “junior consultant,” ranking solution paths by expected ROI, compute cost, and implementation effort.
Serve as a teaching aide, scaffolding learning plans in Blender, MATLAB, or any niche tool a novice needs.

Industry Use-Case Highlights

Autonomous Medical Coding – 30-fold speed-up with human-level accuracy in pilot tests.
Security-Ops Triage – 70 % faster alert classification and enrichment.
Legacy Code Modernization – Generates upgrade roadmaps and unit tests, slashing refactor time by 60 %.
Vendor Due-Diligence – Cross-references filings, news, and technical docs to cut contract-review cycles in half.

Looking Ahead

GPT-5 as a Unified Blend: Rumored to merge O3-style reasoning, GPT-4o’s rapid multimodal generation, and Mini-models’ speed so users no longer juggle model names.
Open-Source Parity: Community-built “DeepSeek R1”-class models may pressure cloud vendors to expose advanced reasoning APIs inside secure HIPAA/GxP enclaves.
Policy & Ethics: As O3 occasionally “reward-hacks” by claiming tool calls it never made, robust audit trails and provenance tags are top research priorities.

Community Q&A

The session closed with a rapid-fire Q&A on memory persistence, pay-walled research, and hardware requirements:

Memory beyond the 200 k tokens likely sits in a transient external store—details still private.
O3 can’t tunnel through pay-walls but finds abstracts and alternative hosts; future open-source agents could accept user credentials for compliant access.
A Mac M-series with 64-128 GB RAM runs multi-billion-parameter local models; Windows users need discrete GPUs or quantized 3 B models.

Stay Connected

Learn More ��ͷ�� AI Flash:
Subscribe for Future Sessions:
Watch the Recording:

The post ChatGPT-O3 Reasoning Agents Unlock Long-Horizon Multimodal Problem Solving appeared first on DSI.

AI Deep Dive: AI for Rapid Response & Restoration

Miguel Nunez — Wed, 23 Apr 2025 18:28:17 +0000

On April 21st, we hosted an AI Deep Dive Session in partnership with Servpro Elite, one of the largest Servpro franchises in Middle Tennessee specializing in fire, water, and mold damage remediation for residential and commercial properties, including large-loss operations for key clients like the U.S. Military. Together, we explored how AI and data science can revolutionize disaster restoration workflows—from job intake and scheduling to damage assessment and resource allocation.

Highlights:

Purpose: Identify opportunities to leverage AI for improving efficiency, accuracy, and customer outcomes in disaster restoration operations.
Focus Areas: Workflow automation for job intake, scheduling, and estimating; image recognition for damage assessment; and predictive analytics for equipment and resource planning.
AI Applications: AI-powered chatbots for client communication, machine learning models for analyzing historical data, and computer vision tools to assist technicians in the field.

Session Insights:

The session highlighted how AI can reduce manual overhead in restoration workflows, enabling faster response times when disasters strike.
Working directly with the Servpro Elite ownership team provided hands-on insight into the real-world challenges of coordinating large-scale restoration projects.
Participants discussed future opportunities for predictive models that could optimize resource allocation and improve decision-making across job outcomes.

Conclusion:

The AI Deep Dive with Servpro Elite showcased how intelligent systems can transform an industry where rapid response is critical to minimizing damage and helping communities recover. This session provided a unique opportunity for those interested in applied AI, operations optimization, and emergency services to engage in meaningful discussion and collaboration.

Are you interested in hosting a future AI Deep Dive? Contact us at datascience@vanderbilt.edu.

The post AI Deep Dive: AI for Rapid Response & Restoration appeared first on DSI.

AI Showcase Spring 2025 Recap

Miguel Nunez — Tue, 22 Apr 2025 13:55:48 +0000

AI Showcase Spring 2025: Where Ideas Became Reality

Event Overview

The ��ͷ�� Data Science Institute opened its doors on April 18 for the Spring 2025 AI Showcase, transforming the 1400 18th Ave S. hub into a living laboratory of ideas. Spanning three buzzing sessions, the afternoon celebrated everything from climate policy analytics to museum-grade interactive art—proof that AI at ��ͷ�� is as interdisciplinary as it is inventive.

Interactive Presentations & Community Engagement

Forget sit-and-listen lectures—this Showcase was a “walk-through of innovation.” Attendees roamed an exhibition floor of hands-on demos, live visualizations, and rapid-fire Q&A, meeting creators one-on-one. The vibe? Part science fair, part startup launch, entirely electric. Industry guests rubbed shoulders with undergrads, faculty mentors brainstormed with alumni, and every table invited “try it yourself” exploration.

Diverse Project Spectrum

From brainwaves to brushstrokes, this semester’s lineup pushed AI into new territory:

Neuroscience & Health – Real-time EEG-to-fMRI translation makes high-resolution neuroimaging more accessible.
Time Management – A mobile planner that optimizes schedules through adaptive learning.
Climate Policy – A platform tracking global legislation and surfacing actionable insights for decision-makers.
Arts & Culture – An interactive Frist Museum partnership exploring identity through AI-generated portraits.
Digital Health – A medication adherence coach that learns from daily habits and nudges users at the right moment.

Winners and Their Innovations

First Place ($1,000) – Deaf Lingo
Lane Burgett (Undergraduate, School of Engineering)
An interactive sign-language platform that uses computer vision to give instant feedback on signing technique, blending accessibility and AI to close the communication gap.

First Runner-Up ($500) – EEG-to-fMRI
Yamin Li (Graduate Student, School of Engineering)
A scalable framework predicting whole-brain activity from EEG signals, poised to democratize high-fidelity neuroimaging.

Second Runner-Up ($200) – MediCheck
Jimmy B. & Haoran Qin (Undergraduates, School of Engineering)
A personal medication advisor that tailors dosage reminders and detects potential interactions—bringing AI to the pillbox.

Community Impact & Industry Engagement

Beyond applause and prize ribbons, the Showcase acted as a networking magnet. Recruiters scouted talent, researchers forged cross-department collaborations, and Nashville tech leaders discovered fresh pipelines for innovation. Many projects began scheduling follow-up pilots before teardown even started.

Looking Forward

If Spring 2025 proved anything, it’s that ��ͷ��’s AI community doesn’t wait for the future—it prototypes it. Expect the Fall 2025 Showcase to be even more daring, with bigger collaborations, deeper industry ties, and a continued focus on real-world impact.

Stay Connected

Don’t miss the next round of breakthroughs. Follow the Data Science Institute on social media, join our mailing list, and mark your calendar for the Fall 2025 AI Showcase. Questions about presenting, sponsoring, or attending? Reach us at datascience@vanderbilt.edu.

The post AI Showcase Spring 2025 Recap appeared first on DSI.

AI Days 2025: A Recap

Miguel Nunez — Tue, 01 Apr 2025 20:36:13 +0000

AI Days 2025: A Recap

March 5–6, 2025 • ��ͷ�� University Student Life Center

AI Days 2025 expanded on the success of our inaugural AI Training Day, growing into a two-day event that explored the full spectrum of artificial intelligence—from foundational concepts to advanced research applications. With over 300 participants joining in person and online, we brought together faculty, students, staff, and industry partners to explore how AI is transforming research, education, and professional practice.

Held at the ��ͷ�� Student Life Center, the event featured multiple concurrent tracks, hands-on workshops, keynote presentations, and our first-ever AI Showcase—an evening of demonstrations highlighting innovative work from across the university.

200+
In-Person Attendees

100+
Virtual Participants

40+
Sessions & Workshops

Thank You to Our Community

This event would not have been possible without the dedication of our speakers, workshop leaders, and organizing teams across the College of Connected Computing, Data Science Institute, and our many partner organizations. A special thanks to everyone who attended, asked questions, and contributed to the collaborative spirit that defined both days.

Day One Highlights: Foundations and Frontiers

The New Age of Models: Jesse Spencer-Smith and Dr. Charreau Bell opened the conference with sessions on reasoning models like DeepSeek R1 and OpenAI o3, and explored how AI agents are shifting from text generation to autonomous task completion.
VAILL Panel – Code and Compliance: Legal experts from across industries discussed emerging AI governance frameworks, offering practical guidance on navigating the evolving regulatory landscape.
Keynote – Ben Hunt on Semantic Signatures: ��ͷ�� alumnus and Epsilon Theory author Ben Hunt delivered a keynote on using generative AI to identify patterns of meaning in historical texts—bridging linguistics, investment analysis, and the humanities.
AI Showcase: The evening concluded with demonstrations from past AI competition winners, Data Science projects, and emerging innovators—a celebration of ��ͷ��’s AI work in action.

Day Two Highlights: Applications and Practice

AI in Protein Science: The Center for Applied AI in Protein Dynamics presented a full track exploring how AI is reshaping drug discovery, from structure prediction to understanding disease mechanisms.
Hands-On Model Training: Umang Chaudhry led participants through end-to-end AI model training, including Parameter-Efficient Fine-Tuning techniques and multi-GPU considerations for scaling.
AI for Nonprofits: A four-part workshop series equipped nonprofit leaders with practical AI skills—from grant writing enhancements to donor engagement strategies.
Fireside Chat – Modern RAG: Curtis Northcutt (CEO of Cleanlab) and Dr. Charreau Bell discussed retrieval-augmented generation and the evolving best practices for building reliable AI systems.

Featured Speakers

A huge thank you to our keynote and featured speakers:

Ben Hunt, CIO at Second Foundation Partners – “Reading the Rhyme of History: Using Generative AI to Identify the Semantic Signatures of Primary Texts”
Curtis Northcutt, CEO of Cleanlab – Fireside Chat on Modern RAG
Jules White, Professor of Computer Science – “Agents Are for Everyone with Amplify AI”
Jesse Spencer-Smith, Chief Data Scientist & Associate Dean, CCC – “The World Changed: AI Overview”
Dr. Charreau Bell, Senior Data Scientist, DSI – “Agents and Actions: A Shift from Text to Tasks”
Myranda Shirk, Senior Data Scientist, DSI – “Energy Consumption of AI Models”
Umang Chaudhry, Data Scientist, DSI – “Hands-On End-to-End Training of AI Models”
Abigail Petulante, Data Scientist, DSI – “Small Models, Powerful Solutions” & “Reinforcement Learning”

Missed the Event? Catch Up Now!

Couldn’t make it to the event? We’ve got you covered. Both days were livestreamed and are now available on our YouTube Channel:

Thank You to Our Sponsors

AI Days 2025 was made possible with support from industry leaders committed to advancing AI education and research:

Oracle • HP • Lenovo • IBM • Microsoft

��ͷ�� Partners

This event brought together major schools, institutes, programs, and offices concerned with AI at ��ͷ��:

College of Connected Computing (CCC) • Data Science Institute (DSI) • School of Engineering (VUSE) • ACCRE • AI Legal Lab (VAILL) • Learning Innovation Incubator (LIVE) • VALIANT • AdvancedED • Center for Applied AI in Protein Dynamics • School of Nursing • Arts & Science AI Grand Challenge

Stay Connected

We’re just getting started. Stay tuned for more workshops and events designed to foster collaboration and innovation in the field of AI. Your engagement is what drives us forward.

If you have any feedback or questions, please don’t hesitate to reach out at datascience@vanderbilt.edu.

Together, We’re Shaping the Future of AI

Our journey continues, and we’re excited to have you with us. Thank you for being a part of our community—and we’ll see you at AI Days 2026!

The post AI Days 2025: A Recap appeared first on DSI.