How MyLÚA Health built a perinatal care agentic platform with IBM watsonx Orchestrate and watsonx.ai
Security-rich, composable solutions help MyLÚA offer a range of AI agents tailored to help birthing parents before, during and after pregnancy.
Pregnancy-related health outcomes hinge on timing. Many of the most serious complications emerge before traditional healthcare workflows detect them, especially during pregnancy and the postpartum period. Supporting parents at these vulnerable times requires systems that are proactive, privacy-preserving and available at the moment of need—not weeks later in a clinical visit.
MyLÚA Health operates at this intersection of healthcare delivery, AI and user experience. The organization focuses on pregnancy and postpartum support, building agentic AI tools that give parents immediate support while coordinating with broader care and financial systems. These systems include doulas, care teams, employers and health plans.
The MyLÚA Health platform is designed to provide birthing parents with emotional, physical, educational and social support. The platform enables caregivers to get a clearer picture of potential areas of need, as it captures emotional, physical and social signals that often go unseen in conventional care models.
A key motivation behind the platform is the preventable nature of many perinatal and postnatal health outcomes. According to the Centers for Disease Control and Prevention (CDC), “more than 80% of pregnancy-related deaths are preventable.” Addressing this gap at scale places stringent demands on data privacy, system reliability and AI governance—constraints that strongly shape the platform’s technical design.
MyLÚA Health built its platform atop IBM Cloud® Code Engine for scalability and simplicity, watsonx Orchestrate® for agent execution and tool invocation and watsonx.ai® for grounded responses.
The core use case is straightforward: deliver timely, privacy-preserving support that fills the gaps between clinical encounters before risks escalate or needs go unmet.
The platform serves multiple constituencies:
- The birthing parent interacts directly with AI agents for education, guidance and resource discovery at any hour.
- Doulas and care teams receive structured insights and conversational summaries that reduce manual documentation while highlighting emerging patterns of need.
- Employers and health plans gain anonymized trend data that helps them understand which benefits are being used, which are not and where outreach might be falling short.
The image presented ahead shows an example of the user interface. The styling imparts a sense of calm and support. Users interact through a chat interface with the MyLÚA Health agent.
For users, value is delivered through immediacy and relevance. For example, a mother struggling with breastfeeding at 4:00 AM can receive evidence-based guidance without waiting for an appointment. A missed postpartum visit due to lack of transportation options can surface previously unknown benefits in real time. These interactions shift care from reactive to preventive while reducing friction across the system.
The system is built around an agentic AI model in which users explicitly choose which agent to engage, rather than relying on fully autonomous orchestration. This design choice keeps decision-making in the hands of the user while still enabling structured automation behind the scenes.
Client applications span mobile and desktop environments, allowing the platform to integrate into existing workflows or operate as a stand-alone app. On the backend, a FastAPI service runs on IBM Cloud Code Engine, a fully managed platform that runs containerized workloads and provides a serverless execution environment to support scalability and operational simplicity.
Data persistence relies on PostgreSQL and Redis, with strict role-based consent and tokenization. Personally identifiable and protected health information is never sent to large language models. This architectural boundary ensures that even if downstream systems are compromised, sensitive data remains protected.
AI workflows are executed with IBM watsonx Orchestrate, which manages agent execution and tool invocation. Reasoning and language generation occur within IBM watsonx.ai. The platform uses retrieval-augmented generation (RAG) backed by curated, evidence-based content across five domains: pregnancy, postpartum care, nursing, mental health and nutrition. Tools include web search capabilities, scheduled background jobs for reminders and check-ins and automatic language detection to support multilingual interactions.
Risk modeling is handled separately through machine learning models that identify early indicators of issues such as depression. These risk signals help customers adjust support strategies without being surfaced directly to users or care teams, maintaining transparency while avoiding unintended clinical interpretation.
Healthcare workloads amplify the cost of architectural missteps. Regulatory compliance, auditability and data isolation are not optional features. IBM’s data, AI and orchestration services align with these constraints out of the box.
Watsonx Orchestrate enables controlled automation without forcing opaque agent behavior. Watsonx.ai supports grounded generation through retrieval rather than open-ended responses, which is critical in wellness and healthcare contexts. IBM Code Engine reduces operational overhead while preserving the ability to scale across tenants and stakeholders.
For users, these choices translate into trust and availability: MyLÚA Health reports that 79% of users reported feeling comfortable sharing sensitive information.
For organizations deploying the platform, they reduce total cost of ownership by minimizing custom compliance work, limiting rearchitecture as usage grows and simplifying integration into existing systems.
By building on IBM’s data, AI and orchestration stack, MyLÚA Health transformed a deeply human healthcare challenge into a scalable, preventive digital platform. The architecture supports growth across new partners, new workflows and new populations without reworking core security or governance assumptions.
For birthing parents, the impact is timely, empathetic support when it matters most. For care teams and organizations, it delivers earlier insight, reduced manual burden and clearer signals about what is working. For the platform itself, IBM’s products provide the stability, compliance posture and flexibility needed to scale preventive care without sacrificing trust or control.