We think Conversational AI 
agents will deliver all professional services 
in India.
We work with best investors
Releases
We're shipping, as always. More exciting stuff soon.
Jan 22, 2026 · 10:56 AMLLMs don't need bigger context windows. They need better context engines. A behind-the-scenes look at how we solved context amnesia and improved multi-turn accuracy from 60% to 90% on our internal benchmark. inside.anuvaya.com/whitepapers/realtime-context-engine
Dec 1, 2025 · 1:28 PM
We're shipping non-stop. Here's some of the stuff we worked on. More soon.
Dec 1, 2025 · 12:14 PMWe started with Astrology.
We're a small group of engineers, designers, and product folks who are building in the intersection of conversational AI & domain expertise. Making an AI agent sound human-like is hard. Making an AI an expert in a domain is also hard. We're trying to do both together.
Our Values
Users ≫ Σ
We speak to our users, have no judgements, and are obsessed with understanding them deeply.
Obsession Over Craft
Obsession is in the details and higher the obsession, more the details seem non‑existent.
First Principles Thinking
We challenge status quo by breaking down problems to their fundamental truths, grounded in logic.
Our Culture
We're not in a rush to hire. We're not even looking for a particular experience or skill. Starting this company was simply a selfish need to look for A players and building a team of them. That takes time.
We care obsessively about craft. Work gets questioned, broken down, rebuilt, because the bar here is higher than most places.
This place will cost you something. You'll think about problems after hours. You'll redo things until they feel right. You'll see teammates who push how you work.
This doesn't work for most people. And we're okay with that, it was never meant to.
Vaya 
Vaya 
Vaya Open Roles
Articles
RCE — Realtime Context Engine
How we feed live, streaming context to agents so they reason on what's happening right now, not what was true ten seconds ago.
Stateful Agent Orchestration
The orchestration layer that lets agents hold state across turns and tools, the way a human consultant naturally would.
Your Eval Is Broken
Most evals measure the wrong thing — and they fail in the exact ways your model does. A note on building evals that actually catch regressions.
FAQs
On the moon (see our astro pun?). We work out of a hacker house in Vasant Kunj. Our current team comes from Gurgaon, Delhi and Noida. We strongly encourage everyone to work from office.
We're huge believers in focus and want to scale our astrology product before we launch another. Broadly, we want to target other professional services next — like general medicine.
Small. We deliberately keep the team tight and the org flat. We're not optimising for headcount.
Not 9-to-5, not managerial, not WFH. We work hard, care about craft, and prefer working together in the same room.