Disclaimer: This summary has been generated by AI. It is experimental, and feedback is welcomed. Please reach out to info@qconlondon.com with any comments or concerns.
The presentation titled Lessons Learned From Shipping AI-Powered Healthcare Products by Clara Matos provides insights into the practical challenges and strategies in deploying AI solutions in healthcare. The session highlights Sword Health's experiences and learnings while emphasizing the importance of balancing safety, consistency, and reliability in AI deployments.
Key Takeaways:
- AI and Healthcare: AI in healthcare facilitates personalized, effective, and scalable care, but faces challenges in regulated environments.
- Developing Guardrails: Essential for maintaining safety and preventing unwanted content, particularly in medical advice and content safety.
- Evaluation and Performance Measurement: Evaluation tools like EOLs and various grading approaches help ensure continued performance without regressions.
- Prompt Engineering: A crucial starting point for optimizing large language model (LLM) performance. Retrieval augmented generation and context optimization further enhance domain specificity.
- Feedback Collection: Emphasizes continuous user feedback to improve system performance.
- Data Analysis and Culture: Promoting a culture of data analysis within teams to constantly identify and address new failure modes.
Overall, the talk outlines actionable strategies for AI development in healthcare, promoting a robust framework of evaluations, feedback integration, and continuous learning for AI systems.
This is the end of the AI-generated content.
This talk provides valuable insights into Sword Health's real-world experience implementing AI in healthcare, focusing on practical strategies for developing consistent, safe and reliable AI-powered healthcare solutions. Through a case study format, attendees will learn concrete approaches to building guardrails, evaluation frameworks, and data-driven development practices. The session will be valuable for those looking to understand the challenges of deploying AI systems in real-world settings.
Interview:
What is the focus of your work?
Building and scaling machine learning systems that support Sword Health's goal of reinventing how patients access and receive care, by creating a more human, more clinically effective, and more scalable way to treat patients.
What’s the motivation for your talk?
I look forward to sharing our practical learnings because, while many discuss AI's potential in healthcare, few are openly sharing strategies for responsible implementation.
Who is your talk for?
The session will be valuable for those looking to understand the challenges of deploying AI systems in real-world settings.
What do you want someone to walk away with from your presentation?
This talk provides valuable insights into Sword Health's real-world experience implementing AI in healthcare, focusing on practical strategies for developing consistent, safe and reliable AI-powered healthcare solutions. Through a case study format, attendees will learn concrete approaches to building guardrails, evaluation frameworks, and data-driven development practices.
What do you think is the next big disruption in software?
How Generative AI will change how we work everyday.
Speaker
Clara Matos
Head of AI Engineering @Sword Health, Focused on Building and Scaling Machine Learning Systems
Clara currently leads AI Engineering at Sword Health, where she is focused on building and scaling machine learning systems that support Sword Health's goal of reinventing how patients access and receive care, by creating a more human, more clinically effective, and more scalable way to treat patients.
Some of the features she has been working on include an AI assistant which supports patients during rehabilitation sessions, a copilot that enhances and scales the Physical Therapist by recommending the best next steps for patient care, automatically filling medical forms based on call transcripts and summarizing key clinical information, among others and a customer support bot available both through chat and phone, which assists patients with all non-clinical tasks.
Find Clara Matos at:
Date
Monday Apr 7 / 10:35AM BST ( 50 minutes )
Location
Windsor (5th Fl.)
Track
Slides
Slides are not available
