Founder-led AI lab · Owned · Partner · Open
Aryabhatta Labs builds owned products like Flair and Vault Breaker, partners with operators on builds like NibbleIQ and Attuned, and authors open protocols like Brand Context Protocol — on-brand guardrailing and compliance for AI agents. Through Vault Breaker, we're also exploring edgeLM — what becomes possible when AI runs fully on your phone.
What we do
We're a small, founder-led lab betting that the next decade of AI is won by teams who can ship real products, partner well with domain experts, and contribute to the open infrastructure agents will run on. Our working question stays the same: how can AI help people and teams move with more clarity, leverage, and control?
Products we build and run end-to-end — from research bet to live users. Flair, Vault Breaker, Brand Context Protocol.
Co-builds with experts and operators whose domain knowledge plus our AI engineering produces something neither of us would ship alone. NibbleIQ with NutriMantra.
Protocols and open-source building blocks we author. Brand Context Protocol — on-brand guardrailing and compliance for AI agents. RAG Reference Template.
Featured
A daily prompt-injection game. Five turns. Five difficulty levels. One secret word the AI is told never to reveal.
The catch: the AI runs fully on-device on your iPhone. No server. No data leaves your phone. It's the first public showcase of edgeLM — our research thread on on-device AI.
What we own
End-to-end ownership — from research bet to live users.
Turn scattered LinkedIn effort into a consistent growth system.
Flair helps professionals show up consistently without treating LinkedIn like a second full-time job. It combines ideation, ghostwriting, profile optimization, and outreach support into one workflow.
"Build your brand. Convert connections."
On-brand guardrailing and compliance for AI agents.
An open protocol that lets brands publish authoritative brand guidance, policy rules, and compliance context that AI systems can read and respect. As agents act on behalf of teams and brands, context needs to travel with the work — portably, machine-readably, and with the brand staying in control.
"Context should travel with the brand."
Partner Builds
We partner with experts and operators when their domain knowledge plus our AI engineering can produce something neither of us would ship alone. Credit and ownership stay with the partner.
"Not a calorie counter · A lifestyle mirror."
Conversational food and lifestyle logging in English and Hinglish. A nutritionist-verified Indian food database. Tracks how food, sleep, and energy connect — and surfaces patterns, not judgment.
"You've got bigger things to think about… wardrobe shouldn't be one of them."
A private menswear styling and wardrobe service. Aryabhatta Labs builds the AI and operational platform behind it — from wardrobe classification at scale, to AI look composition, to generated lookbook imagery, all wired into a stylist workspace and a delivered client lookbook.
Research threads
Privacy by default. Zero API cost. Real-time UX. Open-source small language models (Qwen 2.5, SmolLM2, Phi 3.5 Mini) running natively on phones — no server in the loop. Vault Breaker is where we're learning what becomes possible.
Open Source
We open-source the building blocks we'd want other early teams to have — practical pieces from our stack and workflow that help others learn faster and ship with less reinvention.
A lightweight implementation for Retrieval-Augmented Generation — ideal for learning, prototyping, and real-world experimentation.
How we work
Team & Culture
Small by design. The goal is tight loops between problem, product, and user feedback — not organizational theater.
Founder — 20+ years across HealthTech, EdTech, SaaS, and AI. Previously at Freshworks, HealthKart, TATA 1mg, and SIEMENS.
Get in touch
If you're building with AI, working on a domain problem that could become a partner build, or thinking about brand and policy governance in an agentic world, we'd love to talk.
