India-aware by default

Most AI tools are built for an American or European personal-ops surface. That works in those markets and falls off a cliff in India. Niyra is built in Gurugram, by people who actually have RC renewals and GST filings and EMI reminders to think about. India-awareness isn't a localization layer we'd add later — it's baked into the surface from day one. This post is about what that actually means at the product level. ## Vehicle records, not just calendar entries Most calendar AI in the US handles "remind me about my car insurance next year." Useful. In India, a single car involves four separate expiry dates — RC, insurance, PUC (pollution under control), FASTag balance. The right shape isn't a calendar entry; it's a typed vehicle record with four fields, four expiry triggers, and India-specific cadences (30/14/7 day reminders before each renewal). We ship vehicle records as a first-class type. You photograph the RC, Niyra OCRs it, and the record persists with all four sub-fields tracked. ## Payments — Razorpay first, Stripe second International customers get Stripe; Indian customers get Razorpay. UPI works natively; netbanking works natively; cards work natively. Pricing displays in USD on the site but charges in INR through Razorpay for India buyers at the prevailing rate. No "select your country" awkwardness. For business users, Niyra's Razorpay integration is full-stack — read settlements, create invoice links, refund payments, reconcile orders. The settled-today summary is part of her morning brief if you ask for it. ## Voice — Hinglish, not English-with-an-accent The Indian accent landscape on voice tools is rough. Most STT defaults to American English models that mangle Hinglish — that beautiful, code-switching everyday register that 200+ million Indians actually speak. We use Sarvam as the fallback for Indian English / Hindi, layered behind Deepgram for the general case. The result: she handles "Kal investor call hai, prep mujhe kar do" without choking. Voice synthesis is the same story. ElevenLabs handles English cleanly; we'll be exploring Indian-voice-specific synthesis providers as that ecosystem matures. ## Outbound phone calls — Tata Teleservices When you tell Niyra "call my dentist and book the next opening," the call has to actually land. Indian telephony has its own routing quirks; international VoIP providers don't always handle India landline-to-mobile cleanly. We use Tata Teleservices for outbound calling — the calls land, the audio quality holds, and the call charges are transparent. ## GST + business compliance For founders and consultants, the monthly grind of GST filings, FASTag top-ups, and renewal portals burns hours that nobody enjoys. Niyra's browser automation handles the GSTN portal, FASTag refills, RC renewal, PUC checks — the stuff most people use to spend a Saturday navigating. It's not glamorous AI. It's high-leverage automation for jobs that are genuinely tedious and genuinely valuable. ## Why this matters strategically The under-served market here is real. India is a third of the world by user count of internet-connected smartphones, and the dominant AI personal-assistant tools today either ignore India entirely or treat it as a localization afterthought. The cost structure of running an Indian operation is lower than the equivalent US operation, the integration surface is well-defined (UPI, GSTN, RTO, Razorpay are stable APIs), and the pain points are concrete. We're not trying to win the US market and add India as an afterthought. We're trying to make the best personal AI agent for India and let the design naturally generalize where it does. ## What this doesn't mean We're not building "AI for Indians" in a way that excludes other markets. The same Niyra works for a founder in San Francisco — she just doesn't have to use the FASTag tooling. The features that are India-specific (vehicle records, GST filing, Hinglish voice) co-exist with features that are universal (Gmail triage, calendar booking, memory). But if you're an Indian consultant trying to figure out why your existing AI tools feel slightly wrong for half your daily problems — this is why. The right way to build international AI is to actually live somewhere that isn't San Francisco.