Pukaar.ai: From a Baby's Cry to an Answer in 90 Seconds

A baby cries the same way for hunger, pain, tiredness and discomfort, and a new parent cannot tell which. That gap turns into anxiety, second-guessing and delayed care. Our research found 83% of women experiencing postpartum anxiety and 85% of parents unsure whom to trust for a parenting decision, while roughly 30% of infant deaths could potentially be prevented by early diagnosis. The problem was never a lack of information. It was not knowing which information to believe, at the hour when nobody is awake to ask.

UX research and interaction design across the whole product: strategy, research, information architecture, interaction design, UI and the design system. I worked alongside the CEO, CPO, CTO, Head of AI and more than 10 paediatricians.

We started with 70+ user interviews and a community of 400+ parents asking more than 50 questions a day. That community was the research instrument, not a focus group booked for an afternoon but a live feed of what parents were actually frightened of at the moment they were frightened of it.

A competitive analysis found no direct competitor in India, one globally, and a handful of indirect ones. The gap was real. From the research we built one composite persona, The Anxious First-Timer, and designed for her specifically rather than for an average user who does not exist.

Every screen assumes a parent using it at 2 a.m., one-handed, with a crying baby in the other arm.

That assumption did most of the design work. It ruled out dense layouts, multi-step forms and anything requiring two thumbs. It is why the key flow, what we called the 2 a.m. path, is measured in seconds: from crying to an answer in under 90 seconds. To get there I moved the clarifying questions behind the recording step, so a panicking parent records first and answers later, and kept the escalation options permanently visible on the report screen.

Four decisions held the rest of the product together:

Plain language before medical labels. A parent needs to know what to do, not what it is called.
AI that shows its reasoning. Every insight carries its logic and doctor-verified sourcing, because an unexplained answer about a baby is worth nothing.
Logging stays optional. Quick or detailed, never mandatory, because a product that punishes you for skipping a day gets deleted.
One-handed by default. Reach, tap targets and hierarchy all built around it.

Pukaar.ai shipped with four connected features: a Cry Analyzer, AI-guided infection screening, on-demand expert support (paediatrics, lactation, psychology, nutrition) and verified parent education, on a design system built to carry the longer-term vision of a full AI-powered infant health record.

It is about ending the not-knowing, at the hour when nobody else is awake.

Zero to live in 15 weeks. Before launch it had a waitlist of 154+, 12+ daily active users and 50+ queries a day.

Building the research community before the product: 400+ parents asking real questions daily beat any scheduled study. Designing for one named moment, 2 a.m. one-handed, instead of a generic user. Putting the recording before the questions, because it matched how a panicking parent actually behaves. Making the AI show its reasoning, which is what earned trust from both parents and paediatricians.

Usability-test the 2 a.m. path with genuinely sleep-deprived parents, not rested ones in a calm room. Define the measurement plan before launch so the 90-second target is proven, not just designed for. Involve the paediatricians earlier in wording the AI explanations, because some clinical phrasing had to be rewritten late.