Twitxy Saved

Vikram Vaidyanathan

Verified account This account is verified. Learn more

@VikramV23

Early stage Investor @Z47_vc. Startups + India Edge Believer, Wildlife enthusiast, have to: cycling/basketball vikramvaidyanathan.substack.com
Bengaluru, India z47.com Joined December 2008
1.3K Following 9.4K Followers 2.2K Posts

Latest posts 5

Vikram Vaidyanathan Verified account @VikramV23
Privileged to be leading Confido Health's Series A. Having known Chetan, Vichar and the team for a while, we were impressed with their 15X growth in a year and their impressive word of mouth and win rates vs competition. Their AI agents (Sara, Ryan) now help 1 in 50 Americans, by providing front office staff to US healthcare, and a great example of labour workforce augmentation.

Great to see DeVC, @Together_Fund, Blume Ventures and Justin Saks double down alongside us.

@avnish , @RajatAgarwal167 , @rbalaraman, @assh2win, @RTinkslinger
Chetan Reddy Verified account @mchetanreddy500 Oct 14
We've raised $22.5M in total Series A funding for Confido Health, including a new $12.5M extension led by @z47_vc.

Medicine has added decades to our lives, yet patients still lose weeks to paperwork.

We're building AI agents to give healthcare teams more capacity to care.
View quoted post on X
4 replies 3 reposts 25 likes 1.6K views View on X
Vikram Vaidyanathan Verified account @VikramV23
1/2

How can India play the next phase of AI better? What are our constraints and unique opportunities?

India now has 100 million plus weekly ChatGPT users, with coding queries at 3x the global median and data analysis at 4x, and yet we sit at roughly 76th on per-capita penetration, which means the constraint has never been demand.
0:52
Vikram Vaidyanathan Verified account @VikramV23
1/2

Can you build a chip company from India and sell it to one of the largest semiconductor companies in the world?

Ravi Annavajjhala has done it. Kinara, built out of Hyderabad, was acquired by NXP Semiconductors for $307M.

And that story deserves a victory lap and huge learning for us as an ecosystem! People hear "AI chips" and picture one company in California and a GPU the size of a fridge.

But that's the training side building the brain. Ravi spent seven years on the other half, inference at the edge, where the model does its job on the device itself. His company Kinara designed those chips out of Hyderabad, and NXP Semiconductors bought it for $307M. He explained the whole concept wonderfully with one analogy.
5:42
3 replies 5 reposts 41 likes 5.9K views View on X
Vikram Vaidyanathan Verified account @VikramV23
1/3

Ten years back, every product conversation for Android apps in India began with two questions. How much memory? How big the download?

That constraint is what made Indian mobile work at scale. Spent an hour with Rahul Chari on Intelligent Indians. Same conversation now, just with AI.

Take @PhonePe's AI search. Universal search - any Indian language, any query runs on Azure OpenAI in the cloud. Transaction insights, the one that lets you talk to your own spend history, runs on the edge with Microsoft Foundry.

Chari's design principles are all India-specific choices.
5:45
1 replies 0 reposts 9 likes 2.7K views View on X
Vikram Vaidyanathan Verified account @VikramV23
Everyone keeps asking whether India ends up a data colony in this AI wave.

Spent an hour with Chari on it. Chari to most of us. Co-founder and CTO of @PhonePe. Outside the DPI and NPCI teams, probably the person who's done the most to build the DPI stack. When there's a tectonic shift in technology, he's who I call. So do most people in the ecosystem.

His answer on DPI x AI is that the promise is 100% real - was heartening for me to hear. AI is very good at turning messy unstructured data into structured data. But every time it does that, it loses coherence. The transfer is probabilistic. India crossed that gap a decade ago. Our DPI data is exact, digitised, machine-consumable. Not the noisy raw material AI usually has to reshape.

So the question is what you build on top of it. Rahul Chari push: stop force-fitting a chatbot onto the stack and calling that AI. Get to domain-specific models. Products designed down to the individual instead of the coarse cohorts we've had to design for.

My push on top of Chari's: the data stack is our right to play, not our right to win. We convert it into a right to win only if the intelligence gets democratised and the IP stays Indian.

Foreign capital investing in Indian AI is welcome. But the trade has to be that we democratise access and intelligence to every Indian on the back of it. In Chari's words: it would be a travesty if we hand over this data just to train the world's models.

Tune in for a deep dive into this thread on our Intelligent Indians podcast.

Watch the full video here: youtu.be/BXEbIiBS_Hk
10:23