I am Head of Research at Eigen Labs.
We have raised $250 million+ from a16z (Series B, Strategic), Blockchain Capital (Series A), Ethereal Ventures, and Coinbase Ventures.
Mission
The central principle of my work over the last decade has been building open networks that maximize human agency. In product terms, this takes the form of three properties that everything I have worked on has strived for:
Permissionless
Empower anyone to build and deploy artifacts — software, capital, and the outputs of intelligence — at internet scale, without asking permission.
Attribution
Programmatic credit assignment, so that value generated and captured by those artifacts flows back to the contributions that produced them.
Accountability
Programmatic guardrails against adversarial behavior — a risk that is even more acute in open networks — with penalties proportional to the harm caused.
Products
Products built around this mission, for which I have been one of the core innovators.
2026 —
Autoresearch
Autoresearch is a collaborative, leaderboard-style platform for open scientific research. Domain experts post problem statements backed by an automated verifier that scores correctness and performance, and anyone can point their agents — powered by a Claude or Codex subscription, or a local model — at solving them. What makes it distinctive is collaborative openness: every submission’s solution artifact is visible to all participants in real time, so people build on one another’s work — refining the same approach or leaping to a new one after grasping the solution landscape on the fly — while a live leaderboard turns competition into a reason to understand the subject more deeply. It democratizes frontier research — rapid feedback from an open ecosystem in hours or days rather than months — with domain experts as shepherds who set the problems and steer the community when progress stalls.
Deployments
On
ecdsa.fail,
110 autoresearchers worldwide have pushed open submissions to
49.3% ahead of Google’s published result (as of July 2026).
2025–2026
EigenCompute & EigenAI
EigenCompute & EigenAI are the verifiable compute and inference layers of EigenCloud. EigenCompute runs arbitrary app code and long-running agent logic off-chain while preserving on-chain integrity using TEE-based cryptography, and EigenAI serves LLM inference over an OpenAI-compatible API made verifiable by bit-exact deterministic execution — so any output can be re-run and checked, a mismatch itself proving faulty execution. This lets AI agents managing real assets act on decisions you can prove rather than trust, backed by slashable stake on EigenLayer.
First
in-production bit-exact deterministic LLM inference
2024 —
EIGEN
EIGEN is the universal intersubjective work token: it extends slashing from faults a smart contract can verify on-chain to intersubjective faults — ones no smart contract can verify mathematically, but any honest human observer agrees occurred — enforced through a social-consensus forking.
First
intersubjectively slashable token
2024 —
EigenDA
EigenDA is the hyperscale data availability layer built as the first service secured by EigenLayer restaking. Rollups post their transaction data to EigenDA, which erasure-codes each blob into chunks so that every operator stores only a small fraction of it, while KZG commitments let anyone verify the data is intact without downloading it. Availability is attested by a large operator set backed by slashable restaked stake, giving rollups Ethereum-grade security at a fraction of the cost of posting data on-chain.
Scale
100 MiB/s live throughput — ~450× Ethereum’s native DA · secured by 4M+ restaked ETH at peak
2023 —
EigenLayer
EigenLayer is the coordination protocol that turns staked ETH into shared security for new networks. EigenLayer lets any developer deploy an autonomous software system by borrowing Ethereum’s cryptoeconomic trust instead of bootstrapping its own. Anyone can register as an operator to run these systems: they earn fees for running them, and they deposit digital assets (tokens such as ETH) that can be slashed if they execute the software maliciously.
Scale
$20B+ staked at peak · 50+ autonomous services secured at peak
Education
2011–2016
B.Tech + M.Tech, EE
IIT Bombay, Mumbai
2019–2023
Ph.D., ECE
UW Seattle
Employment
2016–2019
Research Engineer
NEC Labs, Tokyo
2023–2025
Principal Researcher
Eigen Labs
2025–2026
General Manager
Eigen Labs
2026–
Head of Research
Eigen Labs
I am the co-host of the PostAGI podcast, along with Sreeram Kannan.
PostAGI is a podcast on what AGI changes for labour, capital, governance, markets, privacy, and the world. Each season brings together the economists, political scientists, and builders thinking hardest about the world on the other side of the AGI debate.
Episodes
Post-AGISeason 1
-
We Could See State Collapse by the Early 2030s — Samuel Hammond
AGI as an organizational shock: how governments risk obsolescence without rapid reform, and a provocative scenario of state fragmentation in the early 2030s.
July 2026 · 24 min
-
We Already Have a Superintelligence — Allison Duettmann
Applying AI to science now instead of waiting for AGI — open-science experiments, decentralized inference, and how cryptography lets networks be open and private at once.
July 2026 · 16 min
-
AGI makes it all irrelevant — Cremieux Recueil
How China came to lead biotech, whether open science survives as AI labs internalize, and why coordination is what pushes back against AGI’s pull toward concentration.
July 2026 · 33 min
-
UBI isn’t enough. Access to intelligence is. — Emad Mostaque
The end of knowledge work, the case for community-owned AI, and why access to intelligence — not just income — is what people need on the other side of AGI.
June 2026 · 1 hr 9 min
-
The hollow economy — Christian Catalini
As generation costs collapse, verification becomes the real bottleneck of the AI economy — and why cryptographic infrastructure may unlock verifiable abundance rather than a hollow economy.
June 2026 · 1 hr 5 min
-
The productivity boom that leaves workers poorer — Alex Imas
Why “job exposure” misleads almost everyone, which work automation actually targets, and what stays scarce once AI drives generation costs toward zero.
June 2026 · 1 hr 2 min
The CoordinateDistributed Systems · earlier series
-
Market Design and Agentic Commerce: Part 1 — Scott Kominers
How market design principles apply to a world of AI agents that negotiate, transact, and compete on our behalf — covering mechanism design, marketplace dynamics, and what changes when the participants are autonomous agents rather than humans.
March 2026 · 1 hr 3 min
-
Tempo: The Payment-first Stablecoin-native Blockchain — Mallesh Pai
Tempo’s design as a payment-first, stablecoin-native blockchain — its architecture, consensus approach, and what it means to build a chain optimized for payments from the ground up.
February 2026 · 1 hr 6 min
-
Ethereum’s Quantum Leap: Lean Ethereum — Justin Drake
A clean-slate rethink of Ethereum’s consensus, execution, and data availability — covering post-quantum security, fast finality, real-time ZK proving, and L2s.
February 2026 · 1 hr 17 min
-
AlpenGlow: Solana’s Consensus Rewrite — Roger Wattenhofer, Kobi & Quentin Kniep
AlpenGlow redesigns Solana’s consensus and data propagation stack — Rotor for erasure-coded dissemination, Voter for fast/slow path finality, and what it means for validators and MEV.
February 2026 · 52 min
-
The Anti-Framework: Building Chains from Primitives — Patrick O’Grady & Brendan Chou
Commonware’s anti-framework philosophy: compose optimized primitives instead of inheriting a monolithic stack. Deep dive into their Minimmit consensus protocol and upcoming ExoWare.
February 2026 · 58 min
-
Consensus from First Principles — Kartik Nayak
How consensus actually works, how it evolved over 50 years, how today’s protocol families fit together, and what the next frontier looks like — with one of the world’s leading consensus researchers.
January 2026 · 1 hr 52 min