Scaling Personalized Sleep Coaching: Rise Science's Journey with Maxim AI
About Rise Science
Rise Science is a sleep management platform that helps people understand and address the root causes of their energy issues. The platform focuses on two key principles: sleep debt (how much sleep a person owes their body) and circadian rhythm (their body’s natural energy schedule). This science-backed approach enables RISE to provide personalized daily guidance based on users’ sleep and behavior data to help them feel less exhausted and more productive each day.
Rise Science's AI-Powered Coaching Platform
Rise Science recently launched their AI Expert coaching tier, an LLM-powered system that delivers personalized sleep and energy guidance at scale. The system generates insights and recommendations tailored to each user's unique context, delivering them through in-app messages, push notifications, and SMS. It also features character-based coaching, where users are matched with different AI coaches based on their preferences, each with a distinct voice and communication style.
Challenge
Personalized daily guidance has always been core to Rise Science's product. When LLMs opened the door to delivering this at an unprecedented scale through their AI Expert coaching tier, the team faced several workflow challenges.
Their existing development process presented several key areas for optimization:
- Evaluating LLM outputs was entirely manual and time-consuming, requiring extensive human review for every iteration
- The team lacked a systematic process for writing evaluators to assess quality both proactively during development and retroactively in production
- Getting outputs out in a structured way to evaluate them was difficult
- Prompt development required engineering support for each iteration
- Limited visibility into production performance made it difficult to identify and address quality issues as they arose
These challenges limited the team's ability to confidently scale AI-driven features and made it harder to empower product and design teams to contribute directly to AI development.
Rise Science x Maxim Driving AI Quality
After launching their AI Expert coaching tier, Rise Science migrated to Maxim to address challenges in evaluation and iteration. Maxim now powers the development and evaluation workflow for their AI coaching tier, enabling the team to build, test, and scale personalized guidance with confidence.
Rise Science's workflow with Maxim emphasizes reusability, systematic evaluation, and cross-functional collaboration:
Prompt Development with Prompt Partials
Rise Science builds prompts inside Maxim, starting at the instruction level. To scale reliably, the team uses Maxim's prompt partials to create a centralized knowledge base for their AI coaching system.
With prompt partials, the team can:
- Encode their sleep science perspective once and reuse it across all prompts
- Define consistent voice, tone, and communication style across all AI interactions
- Assemble new prompts quickly by combining shared components
This approach reduces duplication and ensures that every AI interaction aligns with Rise Science's product principles. Whenever the team creates a new prompt, they can instantly pull in relevant partials rather than rewriting context from scratch.
Structured Testing and Iteration
Using Maxim's Prompt Playground, Rise Science follows a phased testing approach.
Their workflow typically includes:
- Writing and editing prompt instructions
- Testing with single input-output pairs to optimize core behavior
- Expanding tests across multiple inputs representing different user types and contexts
- Comparing prompt versions and models side by side to test hypotheses
This structured iteration lets product and design teams refine prompts independently, without requiring engineering support for each change.
Quality Assurance with Custom Evaluators
Rise Science uses Maxim's evaluation framework to define quality standards for their AI coaching outputs.
Inside Maxim, the team:
- Writes custom evaluators aligned to their coaching quality standards
- Checks adherence to character voice and communication style
- Evaluates outputs across different scenarios before deployment
- Compares new prompt versions against established quality benchmarks
Evaluators provide fast feedback on changes and help prevent quality regressions as the system evolves.
Production Monitoring and Evaluation
Once prompts are deployed, Rise Science uses Maxim's production logs to monitor real-world AI behavior across in-app guidance, push notifications, and SMS messages.
Using production data in Maxim, the team can:
- Track and review LLM outputs generated across channels
- Attach evaluators to specific traces to assess quality on live outputs
- Identify patterns and edge cases that are difficult to catch in pre-launch testing
Continuous Improvement Loop
Insights from Maxim's evaluations and production logs feed directly back into prompt development. Rise Science uses these insights to refine prompt instructions, update shared partials, improve evaluator logic, and iterate on coaching behavior over time.
This creates a connected workflow where development, evaluation, and production monitoring all happen within Maxim, enabling the team to continuously improve their AI coaching based on real-world performance.
Impact
Maxim has transformed how Rise Science develops and scales their AI coaching features, delivering impact across multiple dimensions of their product development process.
One of the biggest unlocks for Rise Science has been empowering product and design teams to own prompt development. Previously, every prompt change required engineering support. Now, product managers and designers can independently:
- Write and test new prompts
- Compare different approaches and models
- Iterate based on evaluation results
- Deploy improvements without engineering bottlenecks
Maxim democratized prompt development by empowering product and design teams to own the process, which was a huge unlock for us. We can now easily compare models and prompt versions side by side to test hypotheses and drive continuous improvement. This has accelerated our iteration cycles and improved output quality.
- Kellie Maloney, Product Lead, Rise Science
This shift has accelerated development velocity, as the team members closest to user needs can now directly shape the AI experience.
The centralized knowledge base built through prompt partials has delivered significant efficiency gains. Rather than rewriting context or guidelines for each new prompt, the team assembles well-crafted partials that maintain consistency across the prompt library.
Maxim powers our AI coaching tier, enabling us to build, test, and version prompts, evaluate outputs, track production logs, and manage a centralized knowledge base through prompt partials.
- Kellie Maloney, Product Lead, Rise Science
With Maxim's evaluation infrastructure in place, Rise Science has been able to fully realize the potential of LLM-powered personalization. Rise Science's AI Expert coaching now delivers highly personalized guidance tailored to each user's unique context and sleep data.
The systematic evaluation and testing capabilities in Maxim give the team confidence to push the boundaries of personalization. By rigorously testing prompts, comparing outputs, and monitoring production behavior, Rise Science can scale personalized coaching across in-app guidance, push notifications, and SMS.
Beyond the platform capabilities, the Rise Science team has valued the collaborative partnership with the Maxim team.
Having the shared Slack channel has been awesome, and the team has been incredibly responsive. I've been just so impressed with the speed at which the team responds to things, as well as visibility into the process and knowing a ticket's been created, a ticket's being worked on, a ticket's been deployed. That has been incredible.
- Kellie Maloney, Product Lead, Rise Science
This responsiveness has enabled the team to move quickly, get unblocked, and continuously improve their implementation.
Conclusion
Kellie from Rise Science's product team shares three critical lessons for AI Product teams:
- Invest in evaluation infrastructure early. Build robust systems to evaluate LLM outputs from day one. It will save you significant time and headaches as you scale.
- Build reusable prompt partials. Creating well-crafted partials for shared context pays dividends across your entire prompt library and ensures consistency.
- Make prompt engineering a product skill. Product teams should learn and practice prompt engineering rather than treating it as purely a developer responsibility, as it's becoming as essential as any product skill.
Rise Science's approach demonstrates the value of investing in robust evaluation infrastructure early when building AI-powered products. By treating prompt engineering as a core product skill rather than purely a technical responsibility, they've unlocked faster iteration and better outcomes.
As Rise Science continues to expand their AI Expert coaching capabilities, Maxim plays a central role in enabling them to maintain quality while scaling personalized guidance to millions of users. By building the right foundation for AI development, Rise Science is positioned to deliver increasingly sophisticated coaching experiences that help users maximize their energy and reach their potential.
At Maxim, we're committed to helping AI teams build products that users trust and love. If you're looking to scale AI-powered personalization with confidence, let's connect.
Learn more about Maxim here: https://www.getmaxim.ai/
Learn more about Rise Science here: https://www.risescience.com