PregnancyPal - Life Assistant to support pregnant women

An AI assistant powered by Nimble and Bright Data to help pregnant moms plan supportive, personalized days--offering real-time dining, shopping, and medical visit tips based on their daily needs.

Comment

Inspiration

Pregnancy is a life-changing experience that comes with emotional, physical, and logistical challenges. We wanted to create an AI-powered assistant that empowers pregnant women to navigate their daily lives more confidently—making it easier to plan appointments, discover safe places to eat, and find supportive services based on their personal context. Our goal was to blend empathy with AI, using real-time data and intelligent orchestration.

What it does

PregnancyPal is an AI life assistant designed for pregnant women. It: ** Understands free-form requests like: “I’m 33 weeks pregnant, have a doctor’s appointment at 10am, then want to find a brunch spot with pregnancy-safe food.” ** Extracts key intents: schedule constraints, dietary needs, place types, and time/location restrictions. ** Transforms those needs into structured, optimized queries. ** Searches for relevant venues using Nimble’s live Google Maps integration and Bright Data’s booking datasets. ** Delivers recommendations with reviews, safety info, and location context—all in real time.

How we built it

We built PregnancyPal using: ** LangGraph to orchestrate multiple agents in a stateful, decision-based graph. ** Agent 1: A tool-calling agent that interprets user input into structured, actionable prompts. ** Agent 2: A ReAct agent integrated with Nimble APIs to perform real-time venue and review retrieval. ** LangChain, Databricks Model Serving, and Claude Sonnet for conversational reasoning. ** Nimble MCP tools for searching places, filtering by safety, and collecting reviews. ** Databricks notebooks and MLflow for orchestration and logging.

Challenges we ran into

** Agent orchestration: Combining two asynchronous LangGraph agents into one seamless pipeline. ** Tool integration: Nimble MCP’s async structure required careful loading and dynamic runtime handling. ** Prompt structuring: Designing reliable prompts that generalized well across varied real-life pregnancy scenarios. ** Latency: Streaming live reviews and place data while maintaining responsiveness for demo-time interaction.

Accomplishments that we're proud of

** Successfully designed a modular LangGraph agent that parses complex natural inputs into rich, data-driven actions. ** Created a system that offers not just generic chatbot replies but localized, pregnancy-safe recommendations. ** Integrated real-time data sources (Nimble + Bright) with semantic reasoning in under 6 hours.

What we learned

** How to leverage LangGraph's multi-node architecture to isolate parsing and execution logic. ** How to manage async workflows within Databricks safely, especially for tool calling in notebook environments. ** The value of domain-specific assistants—AI is most impactful when it deeply understands a targeted user’s needs.

What's next for PregnancyPal - Life Assistant to support pregnant women

** Integrate calendar syncing for appointment management. ** Add support for wellness check-ins, meditation, and symptom tracking via third-party APIs. ** Expand to postpartum support—including baby-friendly venues and lactation rooms. ** Incorporate feedback loops so the assistant learns user preferences over time.

Built With

Share this project:

Updates