Skip to content
Advait Jayant
Technical founderCSO at OpenGradient

Advait Jayant

I build the trust layer for AI agents before they touch real workflows.

At OpenGradient, I turn decentralized GPU + TEE compute and MemSync into products people can actually buy, trust, and build on. I also write research and produce films.

Thesis Track recordWork
Some hard numbers

Less biography, more signal.

AI infra revenue
7 figures
from zero
hosted models
4,500+
model network
network inferences
2M+
execution volume
verifiable proofs
500K+
signed outputs
campaign views
50M+
technical narrative
01

Thesis

AI agents are moving from chat into real work. The infrastructure around them is still too trust-me.

Operating belief

The winners won't be agents that sound smarter. They'll be systems you can check after they do the work.

That means execution you can verify, memory users can carry, and economic rails that leave a record.

Today
black-box chat
trust me
Next
agent workflows
verify it
  1. 01
    Trust layer

    Make the run checkable.

    If an agent trades, updates memory, or changes access, you should know which model ran, where it ran, and what came back.

    TEE attestation, signed outputs, execution trails
  2. 02
    Memory layer

    Let context move with the user.

    Good agents don't get to forget everything every session, and users shouldn't be stuck in one assistant just to keep their state.

    portable context across models and products
  3. 03
    Settlement layer

    Put real actions on neutral rails.

    When agents touch money, reputation, or access, the record can't depend on the app operator staying honest.

    external commitments for actions that need accountability
02

About

I work on the gap between a cool demo and something people can actually run.

Based
London
Status
UK Global Talent
Education
LBS · BITS Pilani
Now

I'm Chief Strategy Officer at OpenGradient — the Network for Open Intelligence, backed by a16z crypto and Coinbase Ventures.

My lane: product strategy, ecosystem growth, and customer engineering across the GPU + TEE inference network and MemSync.

Product
zero to 7-figure revenue
Network
4,500+ models, 2M+ inferences, 500K+ proofs
Market
1,000+ Korea attendees in one week
Media
50M+ views on OpenGradient films
Before

Before that, I founded SuperSight (later Peri Labs), an Imperial College-anchored AI research lab. The core product was an LLM-powered NL-to-SQL system that reached 200K+ users at 95% accuracy.

Raised $1.5M pre-seed at $30M from Animoca Brands, Blockchain Founders Fund, and Vayner Fund. 30+ enterprise pilots. IP later acquired.

03

Impact

I'm useful when the tech is real, the market is early, and the story is still messy.

2026

OpenGradient

Product strategy, customer engineering, enterprise motion

Turned a new AI infrastructure product from zero into 7-figure revenue.

The network grew across hosted models, inference volume, and signed proofs.

2026

Market entry

Regional GTM, community design, ecosystem ties

Launched OpenGradient in Korea with 1,000+ attendees in one week.

Builder community, local events, and Korean press coverage.

2025

Distribution

Producer, director, writer

Produced a film-led campaign that took OpenGradient past 50M+ views.

A technical thesis became something people could watch and pass around.

2024

Research

First author, market synthesis, technical writing

Published The State of Edge AI with 174K+ launch impressions and academic citations.

Used by builders, investors, and researchers working on edge AI.

2022

Founder track

Founder, product architect, NL-to-SQL pipeline

Built SuperSight / Peri Labs to 200K+ users and 30+ enterprise pilots.

$1.5M pre-seed at $30M; IP later acquired.

Fig_001

Verifiable inference stack

What changes

A model answer is not enough once an agent can move money, memory, or access.

The system has to leave behind a run you can inspect later: what context came in, where the model ran, what it returned, and where the action landed.

Agent run record
01

Run

request / memory / model

Bring the request, user memory, and model execution into one trace.

02

Prove

TEE attestation / signed output

Have the enclave sign what happened, not just describe it.

03

Settle

external record / audit trail

Write high-stakes actions somewhere the operator can't rewrite.

model
open
run
attested
output
signed
state
portable
action
recorded
04

Experience

Product, research, capital, and GTM while the category is still being named.

  1. 2025 — Now

    OpenGradient

    — Chief Strategy Officer

    OpenGradient builds decentralized GPU + TEE infrastructure for hosting, running, and verifying AI models. $9.5M seed led by a16z crypto. My focus is the inference network, MemSync, ecosystem, and products customers can actually use.

    • 7-figure product revenue
    • 4,500+ hosted models
    • 500K+ proofs
  2. 2022 — 2025

    Peri Labs

    — Founder & CEO

    Founded SuperSight (later Peri Labs). Raised $1.5M pre-seed at $30M from Animoca Brands, Blockchain Founders Fund, and Vayner Fund. Selected for the Delphi Labs AI Accelerator(with NEAR) and UT Austin’s incubator. Built the NL-to-SQL pipeline end-to-end to 200K+ users at 95% accuracy. 30+ enterprise pilots. First author of The State of Edge AI. IP acquired.

    • $1.5M raised
    • 200K+ users
    • 30+ pilots
  3. 2019 — 2022

    Technics Publications

    — Technical Author

    Wrote 50+ technical pieces on AI: neural networks, NLP, transfer learning, and big-data infrastructure. Featured in O’Reilly Safari Books Online. 15K+ readers.

    • 50+ publications
    • 15K+ readers
    • O’Reilly Safari
05

Selected work

Research, films, and talks. The common thread: make technical ideas easier to believe, fund, and use.

Campaign still

Making open intelligence feel real.

50M+
WriteDirectLaunch
01·Film campaign

OpenGradient — film work

Producer, director, writer. A launch campaign that made open intelligence feel less abstract.

50M+ views · launch narrative
narrative strategytechnical distributionlaunch surface
Film IFilm IIFilm III
02·Report · 2024

The State of Edge AI

First author. A report on real-time data, privacy, and edge deployment for large models. 174K+ X impressions, 6 academic citations.

174K+ impressions
edge AImarket map
PDFLaunch tweet
03·Report

The AiFi Thesis

A thesis for AI x DeFi: tokenized compute, training data, and model markets.

AI x DeFi thesis
crypto railscompute markets
04·Paper · SSRN

The Economics of Wash Trading

A paper on manufactured volume, incentives, and market structure in crypto.

Market structure
market designincentives
05·Talk

FHE for Consensus in AI Models

A talk on using FHE to coordinate trustless consensus across model outputs.

Verifiable compute
FHEmodel consensus
06

Speaking

Talks and conversations on agents, open intelligence, and verifiable compute.

07Contact

Let's talk.

If you're building AI infrastructure, agent memory, or crypto rails and the trust boundary is getting weird, send the sharp version.

Build

Verifiable AI infra, agent memory, decentralized compute, product strategy.

Think

Research, market maps, crypto rails, AI x capital formation.

Tell

Technical narrative, launch films, talks, and new categories.