Normalized claim
App return rate among assistant users: 3 x increase
Members who interacted with the nutrition assistant were three times more likely to return to the Omada app in general compared to those who did not interact with the tool.
Omada Health scales patient care with fine-tuned Llama 3.1 on Amazon SageMaker AI
Use case typeHealthcare workflow automationUpdated Jan 12, 2026
Omada Health launched an AI-powered nutrition experience on AWS to provide real-time, evidence-based guidance and motivational interviewing to members. The solution used a fine-tuned Llama 3.1 8B model on Amazon SageMaker AI, with Amazon S3 for training data and artifacts and clinician review for safety and quality.
- Organization
- Omada Health
- Industry
- Healthcare
- Location
- United States
- Published
- January 2026
Reported outcomes
3x
app return rate among assistant usersAdoption & scale
4.5 monthsworkflow launch time
Strategic outcomes
Customer experience & trustProvided real-time personalized nutrition educationRisk & complianceMaintained HIPAA-compliant, evidence-based care deliveryCost efficiencyEnabled health coaches to focus on higher-value member interactions
Catalog median for adoption & scale deployments: +250% across 8 reported metrics. Compare benchmarks →
Why do we believe this?Outcome claims, sources, and evidence checks
Normalized claim
Workflow launch time: 4.5 months decrease
Omada launched the entire workflow with the model in 4.5 months.
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- Omada Health
- Provider
- AWS
- Maturity
- Production
- Linked source
- AWS Machine Learning Blog
The trained model was deployed on Amazon SageMaker AI endpoints and used member profile data and conversation history to generate personalized nutrition education in real time
Customer identity supportedSource describes one deploymentMaturity supported
Primary read
Use case focus
Showing 2 of 2
- 1Healthcare workflow automation
- 2Patient engagement
Provide real-time, evidence-based nutrition education and motivational interviewing while preserving clinical accuracy, safety, and HIPAA-compliant handling of member data.
- Omada partnered with AWS and Meta to fine-tune a Llama 3.1 8B model with QLoRA on 1,000 Q&A pairs built from internal care protocols and peer-reviewed literature.
- The trained model was deployed on Amazon SageMaker AI endpoints and used member profile data and conversation history to generate personalized nutrition education in real time.
- Registered dietitians reviewed outputs and feedback loops from LangSmith annotation queues informed prompt updates and future fine-tuning.
- Members using the nutritional assistant were three times more likely to return to the Omada app.
- The workflow was launched in 4.5 months.
- Nutrition question response time dropped from days to seconds.
Architecture
Q&A pairs were uploaded to Amazon S3, where Amazon SageMaker Studio launched a Hugging Face fine-tuning job for Llama 3.1 8B using QLoRA. Model artifacts were stored in Amazon S3. Inference was invoked from the mobile client through a SageMaker AI endpoint using member profile data and conversation history. Outputs were reviewed by registered dietitians and monitored with LangSmith for continuous improvement.
Sources & evidence1
Evidence: Medium50/100Evidence strength
- Customer explicitly identified
- Deployment status explicitly supported
- Quantified outcome available
- Technical implementation details available
Type: Blog PostPublished: Jan 12, 2026Publisher: AWSEvidence: VendorConfidence: Medium
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Customer evidence
Provider evidence
AI patterns
Was this useful?
Community
Comments
No published comments yet.