DecentraRank — Universal AI-Powered Ranking & Indexing Infrastructure for 0G — Requesting $350K Guild on 0G 2.0
Project Name
DecentraRank
Project Website
Coming Soon (the live prototype, live evidence log, and documentation are hosted in the public GitHub repository linked below).
Project Brief
What problem we’re solving
Decentralised networks have made enormous progress on consensus, storage, and trustless execution — but they have not solved the problem of relevance. When a user, application, or AI agent on 0G needs to find the best provider, the most reliable data source, or the highest-quality dataset, they have no native way to do so. They are forced to rely on centralised intermediaries (re-introducing the trust problem decentralisation was meant to solve) or accept whatever returns first.
This is acute for 0G specifically, because 0G is purpose-built for AI workloads. As the network grows, the volume of AI tasks, agent outputs, and on-chain data will far exceed any human’s ability to manually evaluate it. AI agents themselves will need machine-readable ranking signals to make routing decisions at scale. Without a ranking layer, even the best individual applications on 0G cannot deliver on their promise.
Our solution and key features
DecentraRank is a universal, AI-powered ranking and indexing infrastructure built natively on 0G. It is domain-agnostic by design: any application that needs to score, rank, or retrieve data can plug in through a single canonical schema and consume ranked outputs through a unified API.
Four core innovations:
- Universal schema with five domain-agnostic ranking signals (relevance, recency, provenance, reliability, cost), extensible through domain-specific adapters. The same five signals work whether you’re ranking validators, DeFi pools, AI inference providers, or research datasets.
- Multi-Agent Based Simulation (MABS) ranking engine with three agent types (producer, consumer, validator-of-validators) that derives ranking weights from emergent agent behaviour rather than hand-tuning. To our knowledge, this is the first published application of MABS to ranking-signal calibration for decentralised data infrastructure.
- DoE-driven calibration — the MABS calibration loop uses classical Design of Experiments (fractional factorial designs, ANOVA, regression analysis) rather than the grid-search or Bayesian-optimisation heuristics typical in ML-derived simulation practice. This is computationally cheaper, statistically defensible, and produces interpretable variance attribution.
- Native 0G integration across all three modular primitives: 0G Storage for indexed records, 0G DA for index integrity, 0G Compute for verifiable ranking computation.
Target users and market
Every 0G-native application that needs to rank or discover something: DeFi routing engines, AI agent platforms, DeSci research repositories, prediction-market resolvers, on-chain marketplaces, validator-selection tools, content discovery layers. Plus the broader category of decentralised infrastructure beyond 0G — the global enterprise data management market is projected to reach roughly $135B in 2026 and $226B by 2031, and decentralised infrastructure is the next frontier of this category. No production-grade, AI-powered, cross-domain ranking layer currently exists for any decentralised network.
What makes us unique
Five things together that no other team in this space has:
- A live, working prototype today — not a slide deck. 20 passing tests, three-agent MABS engine, calibration loop sweeping ~126 weight candidates, and verified live connection to 0G mainnet (chain ID 16661). All evidence is in the public GitHub repository (linked below).
- A Principal Investigator operationally embedded in the dataset. I run an active 0G mainnet validator at address
0xaED4832042D1204Faf7a97eDD93611A92B20461c, verifiable on 0G ChainScan. The first indexing target is the 0G validator set itself — and I’m transparently included in it. - Direct industry track record of building exactly this class of system. At Ricoh Inc., I designed network infrastructure for ranking and assignment of large-scale commercial print jobs across distributed clients. DecentraRank is the trustless, blockchain-verifiable evolution of that same architectural challenge.
- Methodological depth from academic training. My graduate research at Penn State was on multi-criteria decision-making for resource allocation under uncertainty, conducted under Dr. A. Ravi Ravindran (foundational researcher in multi-criteria optimisation). I also studied Design of Experiments under Dr. Enrique Del Castillo. The combination of MCDM + DoE training is directly load-bearing for what DecentraRank does — the schema is multi-criteria scoring, and the MABS calibration is DoE.
- A complementary engineering co-founder. Kishore Nand (Co-Founder & Backend & Infrastructure Lead) brings 5+ years of production engineering depth across Azure cloud infrastructure, Python, SQL, Angular, and .NET Core, with a Master’s in Computer Science from California Institute of Management and Sciences (USA). Methodologically-grounded design (PI) paired with battle-tested production engineering (Kishore) is the team’s core strength.
How will you integrate 0G?
DecentraRank is not a product on top of 0G; it is foundational infrastructure that uses every layer of 0G’s modular stack as a composable primitive.
Technical implementation details
- 0G Storage — every indexed entity is serialised to canonical JSON, content-hashed for a stable identifier, and persisted on 0G Storage. Updates produce new records that link to predecessors, forming an append-only audit trail.
- 0G DA — the DA layer guarantees that the index is available and verifiable. Anyone can independently reconstruct and audit any historical ranking.
- 0G Compute — ranking computations run with verifiable execution guarantees, ensuring no single party can silently manipulate ranking output. Heavy MABS calibration runs execute on 0G Compute.
- 0G Chain — adapter contracts and schema registries are deployed on 0G’s EVM-compatible execution layer. Live mainnet connection already verified via the ingestion prototype (see the live_run.log file in the GitHub repository).
Which features use which 0G services
| 0G layer | DecentraRank usage |
|---|---|
| Chain | Adapter registry, schema versioning, governance |
| Storage | Indexed entity records, append-only update history |
| DA | Index integrity, reproducible historical rankings |
| Compute | Verifiable ranking computation, MABS calibration runs |
Timeline for integration
9-month plan with four milestones, each with verifiable deliverables and milestone-based funding release:
- M1 (Months 1–2) — Production hardening of the validator-set adapter; schema lock at v1.0; integration tests against live 0G mainnet; on-chain persistence on 0G Storage; Kishore transitions full-time; recruit third engineer. 20% of funding.
- M2 (Months 3–5) — Production three-agent MABS engine deployed; DoE-calibrated weights derived from real 0G data; smart-contract / dApp adapter as second vertical proof of universality. 30%.
- M3 (Months 6–7) — Public REST + GraphQL query API; integration with 0G Compute for verifiable execution; AI inference adapter (third vertical); third-party security audit. 30%.
- M4 (Months 8–9) — Open-source release under Apache 2.0; deployment guide for 0G node operators; research paper draft for peer-review submission. 20%.
Expected benefits from using 0G
- Foundational infrastructure that makes every 0G-native application more useful by providing trustworthy, AI-driven ranking and indexing — reducing duplicate engineering effort across the ecosystem.
- Demonstration of the full 0G stack (Storage + DA + Compute) running together for a real production workload — a useful reference architecture for other infrastructure teams.
- Network economics — DecentraRank will be a sustained, high-volume consumer of 0G Storage, DA, and Compute resources.
- Research credibility — first published application of MABS to ranking-signal calibration for decentralised data, positioning 0G as the substrate for cross-cutting infrastructure research, not just dApps.
Team
- Hanumant Joshi — Principal Investigator. Industrial Engineer (M.Eng., Penn State). Multi-criteria optimisation research under Dr. A. Ravi Ravindran; DoE coursework under Dr. Enrique Del Castillo. Prior production ranking-system work at Ricoh Inc. Active 0G mainnet validator operator. Based in India. GitHub handle: amsquant.
- Kishore Nand — Co-Founder & Backend & Infrastructure Lead. Computer Scientist (M.S., California Institute of Management and Sciences, USA). 5+ years of production engineering across Azure, Python, SQL, Angular, .NET Core. Currently transitioning to full-time on DecentraRank during M1. Based in Bangalore, India. GitHub handle: knpp-byte.
- Third engineer — to be recruited during Milestone 1. Hiring criteria: prior Web3 / distributed-systems experience, Python proficiency, willingness to work full-time on open-source infrastructure.
GitHub Repository
https://github.com/amsquant/decentrarank — public, Apache 2.0 licensed.
The repository contains:
- Full Python prototype with universal schema, signal computation, three-agent MABS engine, calibration loop
- 20 passing unit tests
- The build/live_run.log file — captured log of live connection to 0G mainnet (chain ID 16661)
- The build/validator_index.json file — ranked validator entities under the universal schema
- The build/mabs_convergence.png and build/mabs_calibration.png charts — simulation results
- The SCHEMA.md, INDEX_DECISION.md, README.md design documents
- The docs/ folder — full formal proposal (Word + PDF)
Demo / Prototype
The full reproduction sequence is documented in the repository README. Reviewers can independently verify by cloning the repository and following the commands in the README. The key end-to-end demo runs include:
- The unittest suite (20 passing tests)
- Replay-mode ingestion against synthetic data
- The full MABS simulation with calibration sweep
- Live mode against 0G mainnet (verifies connection to chain ID 16661 and gracefully falls back to replay for validator data when no Tendermint REST gateway is configured)
A short Loom video walkthrough will be added here in the days following this submission.
Documentation
All documentation is in the GitHub repository:
- Full grant proposal in Word and PDF format — in the docs/ folder.
- Universal schema specification — SCHEMA.md at the repository root.
- Indexing target rationale — INDEX_DECISION.md at the repository root.
- Live evidence log — build/live_run.log in the repository.
Social Media
- Twitter / X (PI): hanumant_josh4
- GitHub (PI): amsquant
- GitHub (Co-Founder): knpp-byte
- Email: hanumantjoshi44 at gmail dot com
Hanumant Joshi & Kishore Nand — Co-Founders, DecentraRank — based in India.
This post will be updated with additional links once Discourse trust level permits.