Blog

VISION March 28, 2026 · Malay Baral

Every organization runs on data. But most organizations don't own their data infrastructure — they rent it from cloud providers who store it in jurisdictions they don't control, process it with models they can't inspect, and govern it with policies they didn't write.

AdiOS changes this.

The Three Pillars

Sovereign by Default. Your data stays where you decide. Your models run where you choose. Your compliance is built in, not bolted on.

Intelligence at the Kernel. AI isn't an add-on — it's woven into the operating system itself. From the knowledge graph to the model router, every component is AI-native.

Circular by Design. The platform creates a virtuous cycle: organizations contribute knowledge, the knowledge improves the system, and the system serves the organizations better.

What We're Building

40 components across 5 divisions. A sovereign RDF engine (GraphCore). A CRDT sync protocol (Meridian). Sector-specific ontologies for banking, healthcare, and agriculture. A model router that keeps sensitive inference on-premises.

This is not another SaaS dashboard. This is an operating system for the AI age — and it's being built in India, for the world.

Share this post Link copied!

ARCHITECTURE April 3, 2026 · Malay Baral

I spent significant time building enterprise systems inside and outside India. Rigorous work. But work that served someone else's market.

I came back with a specific intent. To build something foundational. Here. For this market. On our terms.

AdiOS is the result.

The Brain Is the Most Studied System Architecture on Earth

Neuroscientists have spent decades mapping how the brain stores, validates, and promotes knowledge. The core finding is counterintuitive.

The brain does not store facts. It stores confidence-weighted predictions. Every belief you hold carries an implicit probability. When new evidence arrives, the brain updates that probability. This framework is called the Bayesian Brain hypothesis.

Karl Friston formalised this further as Active Inference. The brain is not a passive receiver. It is a prediction engine. It constantly generates hypotheses about the world and updates them when reality does not match.

The architecture consequence: knowledge without a confidence score is not knowledge. It is noise.

The hippocampus encodes short-term episodic memory. Fast. Local. Volatile. The neocortex stores long-term semantic memory. Slow. Shared. Persistent. Between them is a consolidation process. Memories do not simply move from one store to the other. They are validated.

  THE BRAIN                          ADIOS
  ────────────────────────────────────────────────────────

  NEOCORTEX                    ORGANISATION BRAIN
  Long-term semantic memory    Confidence >= 0.95
  Slow. Shared. Persistent.    Curator-governed
            ^                            ^
            | Semantic promotion         | Validation gate
  PREFRONTAL CORTEX            DOMAIN BRAIN
  Executive validation         Confidence >= 0.85
  Peer-reviewed trust          Steward-validated
            ^                            ^
            | Consolidation              | Validation gate
  HIPPOCAMPUS                  PERSONAL BRAIN
  Episodic memory              Confidence >= 0.70
  Fast. Local. Volatile.       User-local

How Intelligence Changes Across Scale

Individual cognition is well understood. You observe. You update. You act. The loop is fast because it involves one agent.

Daniel Wegner identified Transactive Memory. Groups do not store knowledge in one place. They distribute it. Each member specialises. Each member knows not just what they know, but who else knows what.

Scale to a regulated industry. The institutional protocols become machine-executable policy. An RBI guideline. An IRDAI mandate. A DPDP Act provision. Now knowledge carries legal standing.

Consider a credit analyst at an Indian bank. Her personal assessment of a borrower's risk is local and fast. Her team's credit committee adds a validation layer. The bank's risk policy adds a governance layer. The RBI's prudential norms are the final authority. The same question answered at four different confidence levels, under four different governance regimes.

The Extended Mind and Why Software Gets It Wrong

Andy Clark and David Chalmers proposed the Extended Mind thesis in 1998. Cognitive processes do not stop at the skull. When a person uses a notebook, a spreadsheet, or a checklist, those external artifacts become part of the cognitive system.

Most enterprise software treats knowledge as a retrieval problem. You query a database. You get a result. The system does not change.

The brain does not work that way. Every retrieval changes the network. Every interaction strengthens or weakens a connection. This is neuroplasticity at the cognitive level.

Current AI platforms do not compound. They retrieve.

AdiOS is built to compound. Every interaction produces knowledge. That knowledge enters the Personal Brain. When validated, it promotes to the Domain Brain. When curated, it enters the Organisation Brain. The organisation gets measurably smarter with each deployment.

  RETRIEVAL MODEL (current platforms)
  ────────────────────────────────────────────────
  Query -> Vendor database -> Result
  System unchanged. Intelligence does not grow.

  COMPOUND MODEL (AdiOS)
  ────────────────────────────────────────────────
  Interaction -> Personal Brain -> Domain Brain -> Org Brain
                      ^                              |
                      |______ intelligence compounds _|
                              stays sovereign

Why India Needs This Now

India is not a smaller version of a Western AI market. It is a large, regulated, multilingual, sovereignty-conscious economy with its own compliance architecture. DPDP Act 2023. RBI guidelines. IRDAI frameworks. SEBI mandates.

Most AI platforms treat these as edge cases. We treat them as primary design constraints.

Every component knows its jurisdiction. Every data element carries its classification. Every inference call respects the regulatory boundary. Compliance is machine-executable at the OS layer.

Not a checkbox. Not a PDF. Not a third-party audit.

Share this post Link copied!

Malay Baral · Founder, AdiOS Platform Private Limited · Hyderabad, India

adiosplat.io

IDENTITY April 3, 2026 · Malay Baral

India built the world's most powerful identity infrastructure. Aadhaar. UPI. DigiLocker. Account Aggregator. ONDC.

Together they moved 600 million people into the formal economy. They saved the government Rs 4.31 lakh crore in welfare leakage.

That is not a small thing. That is civilization-scale infrastructure.

But there is a gap. It will become a crisis.

AI agents are arriving. Cross-border trade is growing. The DPDP Act demands rights the current architecture cannot fulfill.

The gap is this: India has no portable, user-held, cryptographically verifiable trust layer.

SSI and DIDs fill that gap. Not by replacing what works. By completing what is missing.

INDIA DPI -- THE NUMBERS (2026)
================================================

  IDENTITY
  Aadhaar enrollments       : 144 crore
  Monthly authentications   : 210 crore
  eKYC cost (pre-Aadhaar)   : $10-20 per transaction
  eKYC cost (post-Aadhaar)  : $0.27 per transaction
  Bank account ownership    : 35% (2014) --> 77.5% (2024)

  WELFARE
  DBT leakage savings       : Rs 4.31 lakh crore (2014-2024)

  PAYMENTS
  UPI monthly transactions  : 21.70 billion (Jan 2026)
  Share of global real-time  : 49%

  COMMERCE
  ONDC total transactions   : 350 million+
  Sellers on network        : 1.16 lakh+
  Cities covered            : 630+

The Gap Nobody Talks About

Aadhaar authenticates you. It does not let you carry your proof.

Every institution that needs to verify you must call UIDAI. Every bank, every insurer, every lending app runs its own eKYC transaction. The same Rs 0.27 gets paid hundreds of times for the same person.

Digital loan app conversion rates are as low as 30% on the day of application. KYC friction is the primary drop-off.

India built the cheapest eKYC in the world. Then made people do it dozens of times.

THE MISSING LAYER
================================================

  What Aadhaar gives you:
  "UIDAI says this person exists."
         |
         v
  What you cannot do today:
  "Here is MY proof that I passed KYC.
   Verify it yourself. No server call needed."
         |
         v
  What SSI + DIDs add:
  "Here is a cryptographically signed credential.
   Check the signature. No central server required.
   I choose what to reveal. You get what you need."

Why Blockchain Is the Wrong Answer for India

The DPDP Act 2023 gives citizens the right to erasure. Immutable ledgers directly conflict with this right.

SSI does not need a chain. did:key derives the DID from a public key. Zero infrastructure. did:web serves a DID document from your HTTPS domain. No registry. Revocation uses W3C StatusList2021: flip one bit. DPDP erasure right satisfied.

The Three Futures That Demand SSI

AI Agents on DPI. An AI agent needs to carry your credentials without calling a central server. Current DPI has no agent authorization standard. SSI provides DelegationVC and KYCVerifiedVC.

ONDC Trust Gap. 350 million+ transactions. No identity trust layer between buyer and seller apps. SSI provides GSTRegistrationVC, KYCVerifiedVC, and portable ratings across Beckn apps.

Cross-Border Portability. India has DPI agreements with 23 countries. A Singapore employer cannot verify Indian credentials without calling Indian servers. W3C VCs solve this.

THE DPI EVOLUTION ARC
================================================

  JAM TRINITY (2014-2020)
  Aadhaar + Jan Dhan + Mobile
  Outcome: "The state can see you."
  Rs 4.31 lakh crore saved.
         |
         v
  INDIA STACK (2020-2025)
  UPI + DigiLocker + AA + ONDC
  Outcome: "The state can serve you."
  49% of global real-time payments.
         |
         v
  AI AGENTS + SSI (2025-2030+)
  Aadhaar-anchored VCs + Agent Authorization
  Outcome: "You can act for yourself."
  No central server at verification time.
  Privacy by design. Cross-border portable.

  AdiOS is the enterprise SSI layer
  that connects these three eras.

The Economic Argument

Aadhaar cut eKYC from $10-20 to $0.27. A 97% cost reduction by eliminating physical verification.

SSI cuts further by eliminating repeated institutional verification. Estimated VC verification cost: below $0.01. Check a signature locally. No server call.

The BFSI credit pipeline alone justifies this. Every 10 percentage point increase in loan conversion is billions of rupees unlocked.

The Civilizational Argument

India made its citizens legible. The state could see them, authenticate them, serve them.

SSI makes citizens capable. They act in digital systems without asking a licensed intermediary for permission.

This is not a rejection of India's DPI investment. It is the next layer of it.

India designed UPI and exported it to 10 countries. India designed MOSIP and gave 110 million people digital identity across 20+ nations. India can do the same with SSI. Not by building another blockchain. By building the blockchain-free, DPDP-compliant trust layer the world's emerging economies need.

Share this post Link copied!

Malay Baral · Founder, AdiOS Platform Private Limited · Hyderabad, India