ShardAI

A Volenth AI Brand · Distributed Compute

PROTOTYPE · ONE MACHINE · EARLY-ACCESS INTEREST OPEN

Coordinate idle GPUs for AI compute.

ShardAI is validating a distributed-compute workflow for node operators and AI builders. Verified network metrics will be published as testing progresses.

Register Early-Access Interest How it works

Prototype run, measured

Jobs completed
10 / 10
Verifications passed
2 / 2
Node agents
3 (one machine)
Settlement
Fiat / UPI planned

Figures from our own prototype: a coordinator plus three node agents running on one workstation in Delhi-NCR. One node runs real inference; the other two are simulated. No third-party machines have joined yet.

How it works

Checking work we didn't run ourselves

A machine you don't own returns a result. How do you know it actually ran the job? The coordinator re-submits a share of jobs to a second, independent node and compares the output hashes. Disagreement is detectable without trusting either node.

Dual-submit verification in the ShardAI prototype A job goes from the client to the coordinator, which sends it to one node. A share of jobs is also sent to a second independent node; the coordinator compares the SHA-256 hashes of both results and flags any mismatch. Client submits a job Coordinator job queue node registry + heartbeat verification engine FastAPI + SQLite Node A runs the job Node B independent re-run Node C idle / available SHA-256 compare hashes match → accepted ~20% of jobs every job
Today the coordinator and all three node agents run on one workstation in Delhi-NCR — this is a single-machine prototype, not a live network. The mechanism above is what we are testing.

How it works

Contribute idle GPU capacity

Explore a workflow for contributing spare GPU capacity, with operator terms and payouts still under validation.

Decentralized AI inference

Prototype scheduling and verification for distributed AI inference workloads.

Batch processing at scale

Test whether batch jobs can be coordinated across heterogeneous hardware with measurable reliability.

Distributed compute, under validation

ShardAI is the distributed-compute brand of VOLENTH AI PRIVATE LIMITED. It is prototyping coordination of idle GPU capacity for AI inference and batch workloads.

We are validating node onboarding, job verification, reliability and buyer demand. Early-access interest is open; node counts, capacity, benchmarks and commercial availability will be published only after they are verified.

Founders

  • Amit Gulati

    Amit Gulati

    Founder & Director

    Veteran IT professional — ex international companies & banks.

  • Arav Gulati

    Arav Gulati

    Co-founder & Director

    B.Tech Computer Science, VIT Vellore.

Talk to us

Questions about the prototype, running a node, or early access — write to us directly.

amitgulati@volenth.ai