My Auntie :)

My Auntie is AI postpartum support via call and text because new parents need support, not apps - your caring Auntie at 3 a.m. who listens, remembers, and guides when you need her most.

Comment 7

Inspiration

At 3 a.m., a baby is wailing, your phone is at 4%, and your brain feels like mashed potatoes. You don’t want an app tutorial or a 20-question intake form; you want a human-sounding voice that says, “Hey, love, breathe with me. We’ve got this.” That moment is our spark.

Clinically, the need is huge: Perinatal depression affects around 1 in 7 birthing people, and about 1 in 8 report postpartum depressive symptoms after a recent birth, yet symptoms can appear months later and go unrecognized (Carlson et al., 2025). The combo of stigma and sleep fragmentation makes help-seeking even harder, but they’re treatable when support is accessible.

Plenty of digital tools exist, but new parents often find them overwhelming and feel like they are more work. So we reframed the question from “build another mental-health app” to “what do new parents actually do at 3 a.m.?”. You don’t want an app. You want your Auntie.

What it does

“My Auntie” is a playful, trusted connector that meets you where you are, phone + text, gives one gentle next step, and, when needed, routes you to real care without judgment. It’s Boardy’s low-friction vibe, reimagined for the fourth trimester. Since interrupted sleep can tank mood and cognition like getting only a few hours’ rest, Auntie keeps things tiny, kind, and doable, and yes, she picks up at 3 a.m.

-- For(target users): New parents/caregivers in the first 12 weeks, sleep-deprived, hands full, low bandwidth Especially helpful for first-time parents, C-section recovery, feeding challenges, or limited access to support -- How it helps (fast): Call/Text: zero setup, human-sounding support Triage: Comfort · Resource · Escalate Personal memory: remembers key context to tailor advice Evidence-grounded: tiny, kind, doable steps

How we built it

We built Auntie using a comprehensive tech stack designed for real-time, empathetic support across multiple communication channels. First, we built a phone-first frontend using Twilio to handle SMS/call functionality while the backend leverages Node.js with Express to orchestrate requests and keeps latency low.

Our AI layer strategically combines OpenAI API for fast replies and TTS that powers Auntie's natural voice on calls, while Gemini API handles understanding postpartum concerns in natural language and performs intent-triage to categorize personalized responses as Comfort / Resource / Escalate.

MongoDB Atlas serves as our primary database for storing conversation history and user context, enabling Auntie to remember ongoing chats and build personalized support patterns over time. Snowflake API powers our knowledge retrieval system, allowing instant queries against large datasets of postpartum resources to connect mothers with relevant local support groups, hotlines, and healthcare providers. The frontend system is built with TypeScript to ensure seamless integration between our various APIs, creating a reliable platform that can provide 24/7 evidence-based postpartum care through voice and text.

Challenges we ran into

Building Auntie presented significant technical and logistical hurdles, especially as our first hackathon experience. API Integration proved most challenging, connecting Gemini API for speech and text processing, OpenAI for TTS and fast replies, Snowflake for knowledge retrieval, and MongoDB for memory storage took nearly 18 hours to get working seamlessly together.

Infrastructure Setup was another major obstacle, as we needed a reliable render server deployment since ngrok would have been too slow for real-time voice calls.

Cost Management became a real concern with expenses from OpenAI API usage, purchasing a Twilio phone number, and other service fees adding up quickly.

The Learning Curve was steep, mastering TypeScript and MongoDB integration under tight time constraints while ensuring SMS functionality could remember conversation context and making Twilio calls work smoothly with OpenAI TTS integration.

With our technically demanding project, components frequently broke during development, making the 36-hour time constraint even more pressure-filled as we worked to connect all the moving pieces into a cohesive, reliable platform.

Accomplishments that we're proud of

• Exceeded Ambitious Vision: Started with huge goals and not only met them but surpassed expectations by successfully integrating multiple AI APIs (Gemini, OpenAI) with both SMS and voice call functionality working seamlessly

• Multi-Modal AI Integration: Successfully connected Gemini API to both text and voice channels, enabling Auntie to provide consistent support across communication methods with intelligent intent-triage (Comfort/Resource/Escalate).

• Conversational Memory & Empathy: Built a system where Auntie remembers previous conversations through MongoDB integration and maintains warm, supportive interactions while knowing when to escalate urgent situations.

• Scope Expansion Success: Initially planned for text-only support but successfully expanded to include voice calls with natural TTS, creating a more accessible and comprehensive support system.

• Real-World Impact Potential: Created a technically sophisticated solution addressing genuine postpartum support needs with evidence-based resources and 24/7 availability.

What we learned

We learned how to connect multiple technologies like TypeScript, MongoDB, and different AI APIs (Gemini, OpenAI, Snowflake) to work together smoothly, which was much harder than we expected but taught us valuable troubleshooting skills. Working under a tight 36-hour deadline showed us how important it is to divide tasks effectively and communicate well as a team to get everything done on time. Building an AI system for sensitive postpartum support taught us about creating technology that's both helpful and safe, especially knowing when to escalate serious situations to real professionals. Most importantly, we discovered that combining technical skills with understanding real user needs can help create solutions that actually make a difference in people's lives.

What's next for My Auntie :)

• WhatsApp Integration: Expand communication channels to reach mothers globally through WhatsApp's widespread accessibility, enabling support in regions where SMS costs are prohibitive.

• Multilingual Support: Add language capabilities to serve diverse communities and immigrant mothers, incorporating culturally-sensitive postpartum care practices from different regions.

• Healthcare Provider Dashboard: Build a secure portal for doctors and midwives to monitor patient wellbeing, receive escalation alerts, and integrate Auntie's insights into professional care plans.

• Enhanced Security & Privacy: Integrate advanced encryption and authentication systems to protect sensitive maternal health data and ensure HIPAA-compliant information handling

Sources: Carlson, K., Mughal, S., Azhar, Y., & Siddiqui, W. (2025, January 22). Perinatal depression. StatPearls - NCBI Bookshelf. https://www.ncbi.nlm.nih.gov/books/NBK519070/?utm_source=chatgpt.com

Symptoms of depression among women. (2024, May 15). Reproductive Health. https://www.cdc.gov/reproductive-health/depression/index.html?utm_source

World Health Organization. (n.d.). Mental Health, Brain Health and Substance Use: Maternal Mental Health. WHO. https://www.who.int/teams/mental-health-and-substance-use/promotion-prevention/maternal-mental-health

Built With

+ 28 more
Share this project:

Updates

posted an update —

YEAH AUNTIE!! Live SMS + voice with sub-5s SMS latency, intent triage, and high-quality vetted resources; information-only with deterministic escalation; short-lived context for personalization (full persistence intentionally limited this weekend for safety/reliability). Built from 4 deep listening sessions so Auntie gives tiny, kind, doable steps—especially at 3 a.m.

Log in or sign up for Devpost to join the conversation.