SmallCon 2024: Building Big with Small Language Models
Table of Contents
SmallCon 2024: A Virtual Conference for GenAI Builders
Overview
SmallCon 2024 brings together AI leaders from Meta, Hugging Face, Mistral, Salesforce, Upstage, Nvidia, DoorDash and more to explore the massive potential of small language models (SLMs). The conference focuses on practical insights, best practices, and real-world experiences in deploying SLMs to production.
Vision
It's time to build big with small models! We're exploring cutting-edge trends and practical tips for unlocking value at every stage of the GenAI workflow, from data through deployment.
Key Themes
- Efficiency and performance of small language models
- Enterprise transformation with GenAI
- Production deployment strategies
- Model evaluation and validation
- Synthetic data and continuous fine-tuning
- AI agents and practical applications
Conference Materials
All session materials, slides, and summaries are available in this directory:
Keynotes and Fireside Chats
Technical Sessions
- Transforming Customer Service Analytics with SLMs
- AI Agents that Work: Lessons Learned Building Agentforce
- Data 2.0: Unlocking the Power of Synthetic Data
- Small Models, Big Results: A Look at the Best Performing SLMs
- Supercharge Your SLMs with Automated Continuous Fine-Tuning
- Stop Hallucinating: Best Practices in Model Evaluation
Panel Discussions
Evaluation and Metrics
Schedule
| Time (PT) | Session | Speakers |
|---|---|---|
| 10:00-10:15 AM | Keynote: The Future is Small | Devvret Rishi, Predibase |
| 10:15-10:35 AM | Fireside Chat: Enterprise Transformation | Devvret Rishi, Paul Beswick |
| 10:35-11:00 AM | Tech Talk: Customer Service Analytics | Giuseppe Romagnuolo, Convirza |
| 11:00-11:45 AM | Panel: Future of GenAI | Diego Guerra Orozco, Margaret Jennings, Pavlo Molchanov, Loubna Ben Allal |
| 11:45-12:15 PM | Tech Talk: AI Agents that Work | Manjeet Singh, Salesforce |
| 12:15-1:00 PM | Panel: Productionizing SLMs | Abhishek Patnia, Daniel Hunter, Atin Sanyal, Sudeep Das |
| 1:00-1:25 PM | Demo: Power of Synthetic Data | Maarten Van Segbroeck, Gretel |
| 1:25-1:50 PM | Demo: Best Performing SLMs | Lucy Park, Kasey Roh, Siddharth Ghatti |
| 1:50-2:10 PM | Demo: Continuous Fine-Tuning | Arnav Garg, Predibase |
| 2:10-2:30 PM | Demo: Model Evaluation and Validation | Shreya Rajpal, Guardrails AI |
Key Takeaways
- Enterprise Adoption
- SLMs are becoming increasingly viable for enterprise use
- Focus on specific, well-defined tasks yields better results
- Cost and efficiency advantages are driving adoption
- Technical Innovation
- Continuous fine-tuning improves model performance
- Synthetic data enhances training efficiency
- Evaluation frameworks are maturing
- Production Deployment
- Best practices for SLM deployment are emerging
- Focus on monitoring and evaluation
- Importance of proper guardrails and safety measures
- Future Directions
- Integration with existing enterprise systems
- Enhanced efficiency through specialized architectures
- Improved evaluation and validation methods
Contributing
- Submit issues for corrections or clarifications
- Pull requests welcome for additional resources
- Contact organizers for speaker materials
License
Conference materials are provided under CC BY-NC-SA 4.0 unless otherwise noted.
Contact
- Conference organizers: organizers@smallcon.ai
- Technical support: support@smallcon.ai
- Website: https://smallcon.ai