Sharath Keshava

Building @Sanas_ai, prev: built @observeAI

[Sharath Keshava @sharathk83]

@vickyquests @Sanas_ai Thanks!

[Sharath Keshava @sharathk83]

Today, @Sanas_ai is doing over $75 million in ARR. There was a time when we wanted to return our investors’ money because we just couldn’t get the technology to work.

We had convinced people to back us. Now we had to deliver what we said we could build, and we hadn’t figured out how.

You believe the problem needs solving. You’re less sure you’re going to be the people who solve it. And you’re spending someone else’s money while you find out.

We considered giving it back. We kept working, and eventually, we got it to work.

That last sentence makes it sound much simpler than it was.

Today, I advise dozens of founders. Often, they’ll tell me they’re unsure about something, then almost immediately try to explain it away. As though admitting doubt might make me think less of them.

I understand why. You spend so much time convincing people to join you, fund you and buy from you that saying “I don’t know if we can do this” feels like something you’re not allowed to say.

If you’ve never built a company, that can look like weakness. If you have, you’ve probably had that conversation yourself.

Building a company can feel like chewing glass. I don’t expect someone going through it to sound confident every time we speak.

I want them to be able to tell me what’s actually happening. Otherwise, we’ll spend the whole conversation discussing the version where everything is fine.

I’m proud of where Sanas is today. But I remember when we were the founders having that conversation.

[Sharath Keshava @sharathk83]

We just closed our biggest quarter at @Sanas_ai , was pinching myself for a full week to be sure it’s real :-)

$17M in new ARR, beating our forecast by more than 40%. Sanas has now crossed $77M in ARR! A year ago we went past 25m in ARR, ended the year with 50m & now 77m with ~95% GRR and ~160% NRR.

There is a lot I could say about the quarter. Our existing customers continued to expand with us. More of them are buying multiple products across the Sanas Speech AI platform. We are winning larger enterprises. And we are starting to see entirely new markets open up for us, especially in telco.

Building Sanas has never been a straight line. We have had quarters where everything clicked, and quarters where Shawn and I wondered what we were missing. We have changed products, teams, markets, and probably ourselves more times than I can count.

The one constant has been the people. An incredible group who kept believing we could build something much bigger than where we were at that moment, even when the path there was not obvious.

A year ago, $77M ARR would have sounded insane to me. Today, it feels like another milestone along the way.

To everyone at Sanas, past and present, thank you. Now we march toward $100M ARR, and beyond.

[Sharath Keshava @sharathk83]

@Sanas_ai just made it to No. 8 on the 2026 Inc. 5000 list.

That alone would have made my day. Then I learned we’re also the No. 1 AI & Data company in America and the fastest-growing private company in California.

Six years ago, Sanas was a college project at Stanford. Today, we are one of the 10 fastest-growing private companies in America. That feels surreal to write.

What I love about this recognition is that it measures something very simple: growth. Not hype, funding, or what the market thinks we might become. It looks at what actually happened to the business over the last three years.

And a lot happened.

We went from building one product around accent translation to becoming a Speech AI platform used by more than 200 enterprises across healthcare, telecom, financial services, travel, and retail. Millions of live conversations now run through Sanas every day.

But the part I am proudest of is the people behind that growth.

There were plenty of moments when what we were building was misunderstood. Plenty of hard quarters. Plenty of decisions that looked obvious only after they worked. But our team kept showing up, and so did our customers.

We’ve come a long way in six years, and we’re still very early in what we want Sanas to become.

[Sharath Keshava @sharathk83]

The most dangerous role in your company sounds like the kindest: the manager who protects the team from the founder's urgency. What leaves the top at 100 arrives at the floor at 60, and the missing 40 is where you lost the market. Audit for heat shields. They're always the nicest people you have.

[Sharath Keshava @sharathk83]

By the time a trend is obvious, the returns have already been distributed. You're not early to the gold rush. You're late to the shovel store, and the shovels are marked up 100x revenue.

[Sharath Keshava @sharathk83]

Hire 15 people for 8 people's worth of work and the other 7 will invent meetings.

[Sharath Keshava @sharathk83]

Founders bragging about $0 salaries are advertising one of three things: family money, a lie, or judgment I wouldn't fund. I've backed 40+ companies and never once thought "thank god this founder can't make rent." Desperation doesn't sharpen decisions. It shortens them.

[Sharath Keshava @sharathk83]

https://t.co/Qrc4zim1mn

[Jason ✨👾SaaStr.Ai✨ Lemkin @jasonlk]

https://t.co/d9Wch0HBzh

[Sharath Keshava @sharathk83]

Waiting for perfect market conditions is how you miss every opportunity.
Market's too hot → "I'll wait for correction"
Correction comes → "Too risky, I'll wait for recovery"
Recovery happens → "Too late, I missed it"
There's never a perfect time to build. The founders obsessing over macro aren't shipping.

[Sharath Keshava @sharathk83]

Most CXOs are looking at their roles going extinct. Each role consists of two things:
1. Them being IC - this optimises for being closer to truth, being hands-on, and having a 'taste'
2. Being the spreadsheet operator - forecasts, pipeline, media budgets, sprint planning.

2nd part is getting commoditised with AI. Forward thinking CXOs have accepted that, they've gone back to first principles, but most are resisting AI. They'd have hard time accepting the new reality.

Accept and embrace AI. Become the IC you once were.

[Sharath Keshava @sharathk83]

Founders - the first $1M is on YOU. Not on your VP of Sales, not on your SDR team.
Early customers buy founder's story, not the product. If you outsource this, you're asking someone to sell a story they didn't live. Earn the first million yourself, then hand over the playbook.

[Sharath Keshava @sharathk83]

Once you cross $10M ARR: Net Retention > Net New. Yet I see founders hire A+ Sales VPs and B-tier CS leaders.
Your CS team carries 70% of your ARR, and you haven't really closed the deal until the customer renews.
Post-sale isn't support, it's a second GTM motion.

[Sharath Keshava @sharathk83]

@amarpreetkalkat @Google @Meta @Square @DoorDash @coinbase @gokulr @Sanas_ai You both on iOS and android!

[Sharath Keshava @sharathk83]

When the OG behind @Google, @Meta, @Square, @DoorDash, @coinbase calls you out for ‘building something useful’ you be grateful & thank your stars :-)
Thanks @gokulr for the support. This is very close to our heart and everyone at @Sanas_ai is excited on our launch tomorrow!
@jclin22009 @ScottHickmann

[Gokul Rajaram @gokulr]

I’ve fallen in love with an app that I now use for all my video calls where the other person and I are speaking a different language.

Check out Sanas: real time language translation with video calling!

Everything FaceTime should be. Star Trek’s universal translator IRL.

Link below.

[Sharath Keshava @sharathk83]

Cursor for ML Engineers! Congrats to @withneo team!

[Neo AI @withneo]

Introducing NEO: The first Agentic Machine Learning Engineer.

It works like a full-stack ML engineer that never sleeps:
handling data exploration, feature engineering, training, tuning, deployment, and monitoring, end to end.

Powered by 11 specialized agents, NEO runs autonomously, saving ML engineers thousands of hours and making them 10x faster.

Benchmarks:
Tested on 75 Kaggle competitions, NEO scored a medal in 34% of them, significantly outperforming Microsoft’s RD Agent (22.4%) on the MLE Bench.
This sets a new state of the art for autonomous ML systems.

NEO runs on a novel multi-agent orchestrator, powered by a multi-step reasoning engine, context transfer protocol, and agent memory, built to solve complex workflows end to end.

And with human-in-the-loop mode, you can guide, inspect, and override any step.
You're always in command.

NEO is built for real world workflows and ready for production.

NEO is here to make every ML engineer truly superhuman.

Watch NEO in action:

[Neo AI @withneo]

Introducing NEO: The first Agentic Machine Learning Engineer.

It works like a full-stack ML engineer that never sleeps:
handling data exploration, feature engineering, training, tuning, deployment, and monitoring, end to end.

Powered by 11 specialized agents, NEO runs autonomously, saving ML engineers thousands of hours and making them 10x faster.

Benchmarks:
Tested on 75 Kaggle competitions, NEO scored a medal in 34% of them, significantly outperforming Microsoft’s RD Agent (22.4%) on the MLE Bench.
This sets a new state of the art for autonomous ML systems.

NEO runs on a novel multi-agent orchestrator, powered by a multi-step reasoning engine, context transfer protocol, and agent memory, built to solve complex workflows end to end.

And with human-in-the-loop mode, you can guide, inspect, and override any step.
You're always in command.

NEO is built for real world workflows and ready for production.

NEO is here to make every ML engineer truly superhuman.

Watch NEO in action: