Avi Gaba

Technical founder · AI systemsIndia

I’m a founder and engineer. I run EventHorizon Labs, an independent AI lab where we build AI compute and infrastructure for autonomous agents. From electronics and embedded systems to crypto, and now AI and compute. Off the clock I make music and read way too much physics.

--:--:-- IST

India--:--:-- IST

Based in India, working globally.

I’m Avi. I trained as an electronics engineer, started out in software testing, then wrote climate control software for cars. Since 2023 I’ve run EventHorizon Labs, my independent AI lab, and built with more than twenty teams. Our main product is Singularity Layer, a complete agentic cloud infrastructure with an AI inference network, always-on agents, agentic payments, a fine-tuning lab, our own small decision model and much more.

FounderLive

Singularity Layer

AI compute and agentic infrastructure. Agents and people pay for, run and sell AI services over x402 payments, with a marketplace, no-code Studio, payments skill and on-chain agent registry.

AI compute and agentic infrastructure. Agents and people pay for, run and sell AI services over x402 payments, with a marketplace, no-code Studio, payments skill and on-chain agent registry.
An agentic AI cloud with an inference grid running open models on consumer devices and GPUs, on-demand machines, always-on Agent Pods with their own wallets, encrypted agent backups, and a training lab for synthetic data and fine-tuning. 725,401 requests and 640,237,859 tokens served.
A compact decision model that picks the right paid service for an agent, then hands off to deterministic policy checks before any money moves. 80% complete-flow exact match on a frozen 80-case suite, evaluated once, with weights and prompts fixed before the run.
A mechanism for community-owned, verifiable, confidential AI compute. Each cluster launches a token whose trading fees pay the operators serving it and buy the next GPUs, with payouts settled hourly on chain.
An independent AI lab and engineering studio. Product architecture, smart contracts, backend services, model-integrated applications and infrastructure for more than twenty teams, plus open-source agent tools and MCP integrations.
Built and launched a crypto exchange alone in about 45 hours. It processed more than $5M in trading volume in its first week.
Led development of an AI trading suite on Hyperliquid across agent workflows, product architecture, backend, frontend and infrastructure.
An agentic token launcher and exchange. The project reported more than $25M in trading volume.
GIFU, semantic GIF search. Foliyo, an agent-readable URL for your professional identity. OpenPort, a multi-chain portfolio tracker.

independent nodes connected

Singularity Node, our desktop app

Laptops, desktops and GPUs running our app, serving AI models to the network and earning per request.

725,401

AI requests served

640,237,859

tokens processed

60+

independent nodes connected

15K+

OpenClaw skill installs

years building

20+

teams shipped with

2026

What a 0.8B router taught me about evaluation

From X

2014201620182020202220242026

From electronics to AI

2014 to today

2023 –EventHorizon LabsFounder, independent engineering
2025 – 26EVPlusAISenior developer
2024 – 25TrustSwap · Trust AIProject lead developer
2020 – 22Behr-Hella ThermocontrolSoftware engineer, automotive embedded
2019 – 20C-DAC ACTS PunePG Diploma, Embedded Systems Design
2018 – 19Larsen & Toubro InfotechQA engineer, test automation
2014 – 18UPESB.Tech, Electrical & Electronics Engineering

Publications

Recognition

Gold, Solana × Trends AgentTalentShow hackathon.
All India Rank 45, C-DAC C-CAT national entrance examination.

From metal to models

AI and MLMulti-agent orchestration and tool use, prompt engineering, fine-tuning and evaluating small models, local inference with llama.cpp and MLX, and the infrastructure and tooling that keeps agents running
OnchainSmart contracts on EVM chains and Solana, agentic payments over x402
InfrastructureDistributed inference, GPU and node operations, AWS, serverless on Cloudflare Workers, PostgreSQL, Redis and vector databases
LanguagesTypeScript and JavaScript, Python, C and C++, Solidity, Assembly, MATLAB
HardwareElectronics and PCB design, embedded systems, microcontrollers, microprocessors and SoCs, device drivers, operating systems, CAN bus

Now · 01

Singularity Layer

Agents that run, pay and earn on their own.

Building Singularity Layer, an agentic platform where AI agents can run, pay, earn and get real work done on their own.
Making unlimited AI affordable and accessible to everyone, around the clock.
Building an agent harness that helps any agent research an idea, build it and run it as a real business, end to end, without a human in the loop.
Generative hardware, turning a plain prompt into real, working hardware.

Updated November 2026

Q1
How far can small models go with better architecture, learning, memory and tools, rather than more scale?
Q2
What does a model represent internally at the moment it makes a decision, and can we read it reliably?
Q3
As agents start acting on their own, how do we see what they actually do, steer them, and keep them working safely for people, even once they are smarter than us?

Working on a novel idea, a hard problem in AI, agents or hardware, or something worth building? Let’s talk.

Book a call ivaavi@ehlabs.xyz