Observation

AI’s frontier is crowded—developers race toward agent frameworks and sleek apps, yet India’s long tail, half a billion strong, lingers beyond the smartphone grid. Early efforts like Google’s lightweight Indic models, Sarvam AI’s Shuka 1.0 voice LLM, and AI4Bharat’s IndicWhisper signal promise—voice-driven AI, lean and accessible, is piercing this divide. Basic handsets, not apps, rule here, and in 2025, we see these tools scaling to reach those left behind by silicon’s spotlight.

Why This Is Interesting to Us

While others chase the cutting edge, we’re drawn to the overlooked: India’s vast, disconnected masses. Years ago, AI was cryptic—locked behind screens and literacy walls. Today, mainstream access via voice shifts the game. Lightweight models can now run on minimal tech, delivering utility—think a farmer checking prices in Marathi or a trader logging sales in Bengali, all through a call. OEM tie-ins, like WhatsApp’s integration on basic smart-dumb phones, hint at a path: embedding AI directly into these devices. We predict an outsized impact in 2025—voice AI reaching tens of millions in India’s rural and semi-urban fringes, turning the long tail into a proving ground for scale. It’s a quiet revolution we’re poised to seed.

What We’re Looking For

We’re seeking projects that tackle this frontier:

• Voice-first AI, free of apps or smartphones, built for basic devices.

• LLM-agnostic designs, switching models fluidly by context.

• Self-retraining capability, adapting to localized voice inputs and grassroots nuances unaddressed in initial datasets.

• Utility for India’s long tail—rural, semi-urban, low-literacy users—via OEM-ready solutions.
  
  

Are you working on a project that fits this mold? Reach out to us.

For our latest —investments, insights, career opportunities and more—please subscribe to our bi-annual letter below.