Zarfa

Demo

About

The problem it solves ## The Problem It Solves Every year, **over $200 billion in VAT refunds go unclaimed** because tourists miss flights waiting in refund queues, lose receipts, or face complex paperwork. Similarly, **global payrolls** remain slow and expensive due to high bank fees, compliance hurdles, and multi-day transfers. **Zarfa** solves both problems through a **single on-chain payment infrastructure** built on the **Monad blockchain (EVM-compatible)**. ### For Tourists Zarfa lets travelers **claim VAT refunds instantly** through **MetaMask**, without standing in long airport lines or worrying about missing flights. Refunds are processed securely on-chain, and funds arrive in seconds — **anytime, anywhere**. ### For Employers Zarfa also enables **instant, low-cost global payroll**. Employers can upload a simple CSV, and the system’s **AI assistant** calculates, validates, and executes bulk payments across borders — ensuring **transparent, compliant, and near-instant payouts**. ### Why It’s Better - Eliminates paperwork and manual approvals - Guarantees fast, secure, auditable transactions - Uses **AI + blockchain** for trust, speed, and simplicity In short, **Zarfa makes money move like the internet — instant, borderless, and transparent.**

How it's made

Challenges we ran into ## Challenges I Ran Into While building **Zarfa** and the **Non-Custodial Portfolio Automation Platform**, I faced several technical and architectural hurdles that shaped the final design. One of the biggest challenges was **bridging off-chain AI logic with on-chain execution**. My early design pushed too much logic on-chain, making smart contracts complex and costly to run. I restructured the system into a **hybrid model**, where the AI handles validation and computation off-chain, while only verified actions are executed on-chain — improving both performance and cost efficiency. Another challenge was **gas optimization** during frequent strategy checks that ran every five minutes. The constant on-chain calls initially led to high network fees and congestion. I solved this by **batching transactions** and leveraging **Monad’s parallel execution engine**, significantly reducing gas costs and latency. Ensuring **non-custodial automation** was another hurdle. Automating user strategies without taking control of funds was complex. Integrating **MetaMask DeleGator smart accounts** enabled delegated execution while preserving full user ownership. Lastly, users often entered **vague natural language commands** like “buy when the price drops.” I built a validation layer that clarified intent, processed language context, and converted plain text into structured, executable rules — bridging human language and smart contract logic.