Arnav

Founding Researcher @ Pramaana Labs | Formal Methods Researcher @ UCBerkeley

[Arnav @ArnavAMehta]

Surreal. But building the future always sounds crazy. I'm excited for the world ahead.

[Arnav @ArnavAMehta]

How do I explain my ICML trip to my grandmother?

Robots trained on teens' gaming clips. Kids raising millions to drop out. More investors than professors. AI voices you can't tell from humans.

Half bet on next-word prediction. The other half bet millions against it.

[Arnav @ArnavAMehta]

I find it really hard to convey and pick up intuition from LLMs.

Basically when typing into them I have my lazy writing hat on, and I am just skimming their answers cause most of what is spit out is redundant.

I worry that we will no longer have people with strong intuition

[Arnav @ArnavAMehta]

I’ve honestly always wondered why LLMs are so uniquely terrible at tax. Is saving money just completely out of distribution? 😭

we need formal verification for this

[Pramaana Labs @PramaanaLabs]

Current Frontier LLM models hallucinate with Tax queries on phase-outs, miss quadratic interactions, and leave money on the table; all while sounding confident.

Formal verification of these outputs at run-time is the only meaningful way to address this.
Cell-by-cell 1040 modeling + machine-checked proofs that your tax plan is compliant and provably optimal. Audit-defensible. Fearless.

Authored by: Yoshiki Takashima

Read how we turn complex tax optimization into verifiable truth → https://t.co/XA29ZpaMVV

[Arnav @ArnavAMehta]

@MinseokMatthew Exciting! would love to attend if you have room

[Arnav @ArnavAMehta]

We're obsessed with making models bigger.

The harder problem might be making them waste less. Billions of tokens are spent on work that never creates value. More compute ≠ more progress.

Better utilization might be the next frontier.

[Arnav @ArnavAMehta]

Heading to Seoul for ICML 2026 (July 6–11) 🇰🇷
→ Presenting our Berkeley paper on formally verified co-reasoning
→ Talking Pramaana + meeting folks excited about formally verified AI

Around ICML? Would love to meet. Also got some good after-party recs if you're looking.👀

[Arnav @ArnavAMehta]

Would you trust a doctor that had a 2% chance of lying? In mission critical domains, there is no room for error.

Read our blog about how @PramaanaLabs formalizes medical knowledge to enable provably correct diagnostic reasoning.

[Pramaana Labs @PramaanaLabs]

Medical AI is rushing toward autonomous agents. But in the clinic, a fluent reasoning trace isn’t enough.

We need proofs that are machine-checked and respects every inclusion, exclusion, and contradiction in the patient’s data.

Introducing the architecture of clinical truth: formalizing diagnostic criteria in Lean 4 so every diagnosis is verifiable. No more unsafe shortcuts.

Authored by: @kaush_ality @ChristineTataru

Read more about it here → https://t.co/E5eIvAtDnY

[Arnav @ArnavAMehta]

A German court reportedly held Google liable for false AI-generated statements.

Can insurers, banks, and hospitals be accountable for AI decisions they can't verify?

The future of AI isn't just smarter, it's provably correct.

https://t.co/4CPUgEcrio

[Arnav @ArnavAMehta]

Proof should not be confined to mathematics.

LLMs have created an abundance of generated knowledge. But with so much slop going around proof is the only way we can reliably distinguish what is true.

We're bringing proofs to the everyday.
Go read our latest blog!!

[Pramaana Labs @PramaanaLabs]

Math proofs are cool, but the real revolution?
Verifying real-world claims in accounting, tax law, compliance, medicine & more; reliably, efficiently, and at scale.

Proof trees look different across domains. One-size-fits-all won't cut it. AI prover architectures must be purpose-built to navigate the specific reasoning trees, resource constraints, and rule stability of the environment they are verifying.

Read why specialized provers are the future → https://t.co/FbNiJ3Knhy

Authored by: @ArnavAMehta

Watch this space for more to come from our Technical blog-post series.

[Krishnan Raghavan @krishnan_rag]

Thank you very much for the shoutout @svembu! This means a lot and we have been going deep and building domain-specific programming languages to capture strict nuances in areas like law and tax. Exciting times ahead!

PS: I was sitting right next to @sanjaygsub during your inspiring talk at IITM :)

[Sanjay Ganapathy @sanjaygsub]

Sridhar Vembu received the distinguished alum award at IIT Madras in 2016, the year I graduated. I distinctly remember seeing him for the first time, giving a talk on entrepreneurship to our graduating CS batch.

A decade on, to hear him give a shout out to @PramaanaLabs and our thesis on national TV is definitely a goosebumps moment! Thank you @svembu 🙏 Let's bring efficient, verified intelligence to the world!

[Kanu Gulati | Partner @ Khosla Ventures @KanuGulati]

We are thrilled to back @PramaanaLabs!

As AI takes on more important work, trust can’t depend on a human checking every answer. Pramaana is building the capability to make AI outputs verifiable by design; this is a BIG step toward safer, more capable, and more useful AI.

And with the right leadership in @ranjan_vittal, @krishnan_rag, @sanjaygsub and team @PramaanaLabs

cc @khoslaventures @vkhosla

[ranjan_raj @ranjan_vittal]

Today, I'm thrilled to announce Pramaana's $27M seed, led by @khoslaventures.
The foundational domains that hold the world together: tax, law, finance, healthcare; all run on certainty. Probabilistic AI can't give them that. We’ve been asked to accept wrong answers with AI as ‘hallucinations’, while in traditional software terms, it’s just a bug. And a wrong answer in such mission-critical domains is more than just a bug, it's a liability that could have catastrophic impact.

We built Pramaana to deliver a 100% trustable experience to the domains that run on certainty: AI that is provably correct, not probabilistically correct. We turn statute and regulation into machine-verifiable code, so every output ships with mathematical proof of correctness. Our mission is to make AI take ownership of it’s work.
Pramaana in Sanskrit stands for “means of valid knowledge”, and we’re going to achieve that by formalizing the world’s knowledge.

[Arnav @ArnavAMehta]

Formalization enables coordination across multiple constraints simultaneously. Making assumptions explicit helps both humans and AI reason about tradeoffs and allocate resources more effectively. Resonates with ideas @vkhosla often discusses.

[ranjan_raj @ranjan_vittal]

Last week @vkhosla anchored our summit. Injured, when stepping back would've been expected, he showed up and turned it into an example of how verification can transform healthcare. The conviction he brings to every founder, no matter how small the company, is humbling. The GOAT.

[Arnav @ArnavAMehta]

Compilers are the feedback loop LLMs need both at both training and inference. We have scaled it for coding, let’s scale it for real-world domains!!

[ranjan_raj @ranjan_vittal]

@TechCrunch I’m sure our underlying thesis would excite most.
Read more about why we're building this, and how our Domain Formalizer + Auto-formalizer turns regulatory text into verifiable code:
https://t.co/lEAf731rLb

[Arnav @ArnavAMehta]

The first time @krishnan_rag pitched Pramaana to me, his deep passion and unwaivering spirit to solve really hard problems is what drew me in!! The worlds hardest problems are not unsolvable they are just unformalized 🚀🚀

[Krishnan Raghavan @krishnan_rag]

https://t.co/f4ch14xQiX

[Nexus Venture Partners @NexusVP]

The next frontier in AI isn't speed or scale. It's proof.

@PramaanaLabs is building the accountability infrastructure for the world's most consequential domains.

🔗 https://t.co/P0iWQ7KQOP

[Arnav @ArnavAMehta]

@TechCrunch say 'no' to hallucinations!

[TechCrunch @TechCrunch]

Pramaana Labs raises $27M seed round from Khosla Ventures to bring formal verification to AI https://t.co/b2ZbyHjYSx