I am PAWAN UPADHYAY, an independent Web3 and AI builder focused on turning emerging protocol primitives into usable products.

My work spans agentic applications, programmable money, stablecoin infrastructure, on-chain intelligence, privacy-preserving computation, wallets, identity, cross-chain value movement, and AI-native developer experiences.

I prefer building systems that are measurable, verifiable, security-conscious, and useful beyond a single demo.

Veridex is my flagship current build for Telegraph: a deterministic smart-contract intelligence layer designed to produce auditable, machine-readable signals for agents and applications.

The architecture emphasizes:

• proxy and implementation resolution
• verified ABI preference
• independent deterministic checks
• evidence provenance
• confidence and explicit failure semantics
• observable fallbacks
• separation between analysis logic and protocol integration
• security and fraud-oriented contract intelligence

The objective is not to make an LLM guess whether a contract is safe. It is to make meaningful conclusions traceable back to the observations that produced them.

→ Veridex

Building and experimenting deeply around the Arc ecosystem and Circle's programmable-money infrastructure, including:

• ARCTIS — an AI operating environment combining AI OS, Knowledge OS, DeFi OS and Economic Agent OS.
• ArcVault — mobile-first wallet architecture focused on agent-ready financial workflows, identity routing and programmable jobs.
• ArcFlow — intelligent value-routing concepts for simplifying cross-chain transfer and bridge decisions.
• Circle rails, USDC, CCTP-oriented bridging and wallet-controlled transaction flows.
• Agent permissions, budget-aware execution and the human approval boundary: Propose → Review → Approve → Execute.

ARCTIS is designed around a core thesis: AI should be able to reason about economic actions without silently taking custody of a user's money.

→ ARCTIS · ArcVault · ArcFlow

Explored Zama / FHEVM through privacy-oriented smart-contract work, including:

• Private Vote — an FHEVM-based private voting dApp prototype.
• Zama DCA Bot — privacy-preserving DCA logic with batching using FHE/FHEVM concepts.
• Additional Zama builder experiments and test repositories.

→ Private Vote · Zama DCA Bot

Exploring the intersection of AI agents, cryptographic verification and decentralized infrastructure through Soundness-oriented development and ecosystem experimentation.

The broader direction of my work is to make AI systems more accountable: agents should not only act, but increasingly be able to provide evidence for what they observed, decided and executed.

→ Soundness

Built Autheo Wallet, a lightweight EVM wallet engineered for the Autheo Testnet with:

• universal injected-wallet support
• real-time balance synchronization
• live transaction ledger history
• transaction confirmation timing and gas tracking
• dynamic receiving QR codes
• voice-confirmed transaction feedback
• mobile-first serverless deployment

→ Autheo Wallet

Worked on builder identity and reputation primitives, including a GIWA Builder Passport with an on-chain soulbound identity model, public profiles, skill metadata, mint/edit flows and transaction receipts.

→ BuilderPass

Built FactAnchor, a trust-oriented Intelligent Contract experiment for GenLayer that resolves natural-language claims through independently evaluated web evidence and validator consensus.

This work reflects a recurring design theme in my projects: AI reasoning becomes more useful when its evidence, provenance and failure modes are explicit.

→ FactAnchor

Active ecosystem work around Billions Network, with a focus on decentralized identity, human verification, reputation and incentive design.

My interest in this area connects directly to the larger thesis behind my product work: agentic economies need reliable identity and human accountability layers, not only autonomous execution.

Recent projects and prototypes include ProofLoop, ActionPermit, Repair Atlas, ArcVault Wallet, ArcFlow Router, ARCTIS, Base Utility Hub, Monad Pulse, BuilderPass, FactAnchor, Axon402, Verified Agent Identity, and related experiments.

These projects have expanded my practical experience across agent governance, browser verification, on-chain intelligence, wallet security, authorization systems, identity, x402, cloud deployment and production-oriented testing.

A major focus of my current work is the transition from wallet-centric Web3 → agent-enabled economic systems.

Human intent
     ↓
AI reasoning
     ↓
Deterministic policy / risk checks
     ↓
Agent proposal
     ↓
Human or policy approval
     ↓
Programmable execution
     ↓
On-chain proof + history

The important distinction is between an AI that can suggest an economic action and an agent that can execute one safely.

My preferred architecture keeps:

• user-controlled keys and assets
• explicit spending budgets
• deterministic financial intent parsing
• observable execution history
• auditable decisions
• human approval where appropriate
• clear failure states

This is the foundation I am using to explore the Agentic Economy as a real product category rather than simply an AI chatbot connected to a wallet.

Frontend: React, Vite, TypeScript, Next.js, Tailwind CSS, responsive/PWA interfaces
Web3: Solidity, EVM, Viem, Wagmi, RainbowKit, smart-contract integration
AI: Google ADK, Gemini, LLM routing, structured intent parsing, agent orchestration and evaluation
Backend/Data: Node.js, Python, Firebase/Firestore, CockroachDB, REST/API development
Payments: USDC, programmable money, CCTP-oriented cross-chain flows, x402
Privacy: FHE/FHEVM concepts and privacy-preserving smart contracts
Security: Firebase ID-token verification, jose, authorization, risk scoring, human approval and fail-closed execution
Verification: Kane CLI, headless browser testing, deterministic checks, evidence pipelines and production smoke tests
Deployment: GitHub Actions, Vercel, Google Cloud Run, AWS environments and cloud terminal workflows

Cloud Terminal · Amazon/AWS Terminal · CockroachDB · GitHub Actions · Vercel · Google Cloud Run · Kane CLI · Firebase · Google ADK · Gemini · Git/GitHub

Build the system, not the demo.

• Evidence over hype — claims should be traceable to implementation or verifiable sources.
• Deterministic where money is involved — financial execution should not depend on unconstrained model output.
• User controls value — agents should not silently take custody of user funds.
• Fail visibly — infrastructure failures should never masquerade as successful execution.
• Security boundaries are product features — permissions and approval flows belong in the architecture.
• Composable primitives — build components that other agents, applications and protocols can reuse.
• Real UX matters — blockchain infrastructure becomes valuable when normal users can actually use it.

AI
 +
Agents
 +
Identity
 +
Privacy
 +
Programmable Money
 +
On-chain Verification
        ↓
   AGENTIC ECONOMY

I am particularly interested in building the infrastructure and applications that allow humans, AI agents and programmable financial systems to interact safely and verifiably.

• GitHub: @pawansatoshi
• X: @pawansatoshix
• YouTube: Pawan Satoshi
• Telegram: Pawan Satoshi
• Email: pawansatoshi@gmail.com