An Initiative by BALAJI IT Solutions
India's citizen-powered sovereign inference engine.
DUM360 pools India's already-paid-for idle compute — phones, gaming PCs, university & PSU labs, MSME machines, and solar GPU rigs — into the country's largest sovereign AI-inference engine. Built for inference and parallel workloads, not frontier training. Earn via UPI. Indian-hosted platform, DPDP-aligned.
Current stage: Waitlist + Landing Page. The platform (node agent, orchestrator, inference engine) is in design. This repo is where it's being built — publicly, from the ground up.
| Directory/File | What it is |
|---|---|
index.html |
Landing page at dum360.com — the public front door |
portal.html |
Authenticated contributor portal (view/manage your signup) |
privacy.html, terms.html, cookies.html |
Legal pages |
firebase-config.js |
Firebase SDK bridge — signup writes to Firestore, Google Auth |
firestore.rules |
Firestore security rules (PII-protected, one-entry-per-account) |
firebase.json |
Firebase Hosting config + security headers |
scripts/setup-firebase.sh |
Automated Firebase project setup |
docs/PRD.md |
Product Requirements Document — the full platform spec |
docs/compute-benchmark.md |
Honest benchmark — where DUM360 would stand in global compute |
docs/partnership-note.md |
University partner outreach template |
docs/website/PRD.md |
Website PRD — content, data model, rollout phases |
docs/website/DESIGN_SPEC.md |
Design & UX specification |
docs/website/DEPLOY.md |
Deploy runbook (Firebase Hosting + Firestore) |
docs/website/legal/ |
Legal copy source-of-truth (Privacy, Terms, Cookies) |
DUM360 treats every powered-on, under-utilized processor in India as latent national infrastructure. Rather than building new data centres, it harvests capacity that already exists:
| Source | Typical idle window | Why it matters |
|---|---|---|
| Consumer phones & PCs | Nights / off-hours | Largest device count; burst throughput |
| University & research labs (IIT/NIT/IIIT) | Evenings, weekends, vacations | LAN-clustered + NKN-linked → real parallelism |
| Government & PSU labs | Off-hours, spare capacity | High trust, sovereign by default |
| Enterprise / MSME / startup machines | Outside business hours | Wired, powered, predictable |
| Solar GPU rigs (Gujarat / Rajasthan) | Daytime surplus generation | Cheapest, greenest cycles |
Dual-use mandate: Commercial peace-time mode (affordable AI inference for Indian startups) + Rashtra Seva emergency mode (national strategic compute reserve for disaster response).
For the full vision, architecture, and functional requirements, read docs/PRD.md.
We don't claim to beat hyperscalers. Read the full analysis in docs/compute-benchmark.md. The honest bottom line:
DUM360 could plausibly deliver a few exaFLOP/s of usable AI-inference throughput — 1–2 orders of magnitude beyond India's entire current public supercomputing capacity — without building a single new data centre. It is built for inference and parallel workloads, not frontier model training, which consumer networks physically cannot support.
- Landing page live at dum360.com
- Contributor sign-up flow (Firebase Firestore + Google Auth)
- Contributor portal (view/edit/withdraw your signup)
- Legal page drafts (pending counsel review)
- Alpha desktop app — K3s-based node agent for 1,000 student/dev machines
- Central orchestration layer on Indian cloud provider
- First distributed workload (100 concurrent nodes, no data loss)
See ROADMAP.md for the full phased plan.
We need builders across the stack — distributed systems, WASM/K3s edge agents, secure enclaves, UPI integration, and more. This is an early-stage project, which means early contributors shape the architecture.
Read CONTRIBUTING.md for how to get started.
- Live site: dum360.com
- Sign up: dum360.com/#join
- Honest benchmark:
docs/compute-benchmark.md - Platform PRD:
docs/PRD.md
MIT — see LICENSE.
Made in India 🇮🇳