VELT - AI Powered DeFi Lending

Project Name

VELT AI – Conversational DeFi Lending on Hyperion


Problem Statement

Decentralized lending platforms are often intimidating and inaccessible to everyday users due to complex user interfaces, confusing workflows, and a steep learning curve around collateral management. Many potential users are excluded simply because they lack the technical skills or confidence to interact with DeFi protocols directly.


Solution Overview

VELT AI transforms decentralized lending into a simple, chat-based experience. Powered by the Alith SDK and built on the Hyperion L2, users can lend, borrow, and manage their loans through natural language interactions. Whether via a Telegram bot or a web-based chat interface, users gain intuitive access to DeFi without navigating complicated UIs. Hyperion’s speed and low fees make it the ideal execution layer, while Alith bridges user intent to on-chain actions.


Project Description

VELT AI integrates advanced AI with robust DeFi infrastructure to offer a conversational interface for crypto lending.

Core Features

  • Natural Language Lending – Interact with smart contracts using simple phrases like “Lend 10 USDT.”
  • Hyperion-Powered Speed – Enjoy seamless transactions thanks to parallel execution and sub-second finality.
  • On-Chain Credit Scoring – Borrowing power is assessed based on wallet activity and protocol engagement.
  • Real-Time Risk Alerts – Users are notified if their collateral ratio approaches liquidation thresholds.
  • Safe, Guided Transactions – Transaction previews ensure clarity and require explicit wallet confirmation.

Technology Stack

  • Alith SDK – Natural language intent parsing
  • Smart Contracts – Deployed on Hyperion
  • Frontend – Built with React + Next.js
  • Telegram Bot API – For conversational DeFi interactions
  • On-Chain Modules – Power credit scoring and risk tracking

Technical Details

Intent Parsing

Using Alith’s SDK, we extract user intent from commands like “Lend 10 USDT.” These inputs are processed by a middleware layer that maps them to verified smart contract calls via ABIs and pre-defined schemas. Every action is preceded by a confirmation summary — e.g., “You’re about to lend 10 USDT. Confirm?” — to ensure user clarity and control. No transaction executes without a wallet signature.

Why Hyperion?

Hyperion provides:

  • Sub-second finality
  • Parallel execution
  • Low fees

These advantages enable real-time interactions that feel conversational and responsive — essential for chat-based DeFi.

Translating DeFi Complexity

Advanced DeFi concepts (like liquidation thresholds or APY) are simplified through a backend module that tracks on-chain metrics and delivers concise summaries. Examples:

  • “Your collateral ratio is 75%. Liquidation occurs at 80%.”
  • “If ETH drops by 15%, you risk liquidation.”

The Alith AI communicates these insights clearly, while the core logic is managed on-chain or through verified indexers.

Handling Ambiguity

If user intent is unclear (e.g., “Take some ETH”), the AI requests clarification before proceeding. Every action is previewed and requires explicit wallet approval — ensuring safety in uncertain contexts.

Differentiation from Existing DeFi Platforms

VELT AI is not just a chat-based version of existing protocols like Aave. Its core innovations include:

  • On-chain credit scoring for smarter, safer borrowing
  • Conversational automation to guide beginners step-by-step
  • Real-time alerts on risk metrics via the chat interface
  • Hyperion-native execution for unmatched responsiveness

Together, these features deliver a safer and more accessible experience — particularly for users underserved by traditional DeFi platforms.


Hackathon Focus: HyperHack

We’ll showcase:

  • Seamless intent-to-transaction flow with Alith
  • Lightning-fast lending and borrowing on Hyperion
  • Real-time health checks and risk alerts
  • Transparent credit scoring logic, explained in plain language

Community Engagement Features

To gamify onboarding and incentivize participation, we’ll implement the following:

Testable Features

  • Deposit into the lending pool
  • Borrow against collateral
  • Use AI chat to perform DeFi actions
  • Check loan health/collateral ratio in natural language

Points System

  • +10 points – First successful deposit
  • +15 points – Borrow using AI chat
  • +5 points – Use the health monitoring feature
  • +20 points – Invite a friend to try the chatbot

Gamified Experience

  • Users level up by completing tasks
  • AI chat offers hints, tips, and surprise challenges
  • Leaderboard encourages competition and sharing

User Onboarding

  • Chat-first experience removes technical friction
  • Perfect for crypto-curious or DeFi newcomers

Getting Involved

  • GitHub: Codebase to be made public after initial demo
  • Community: Telegram and Discord channels launching during hackathon week
  • Team Contacts:
14 Likes
  1. Is VELT AI building its own lending protocol or integrating existing ones (e.g., Aave, Compound)?
  • If custom, how is liquidity bootstrapped on Hyperion?
  1. Will users be able to customize loan terms like duration, rate type (fixed/variable), or is it standardized for simplicity?
  2. How is the credit score calculated?
4 Likes

Your proposal caught my attention, and I have a few questions:

:backhand_index_pointing_right: How does the on-chain credit scoring model evaluate wallet activity to determine borrowing power while maintaining user privacy?
:backhand_index_pointing_right: What safety measures are in place to prevent accidental or malicious lending/borrowing actions through conversational misunderstandings?
:backhand_index_pointing_right: Can users customize or opt out of real-time risk alerts, and how are those alerts delivered across different interfaces?
:backhand_index_pointing_right: How scalable is the system in handling multiple simultaneous users interacting via chatbots on Telegram or web without latency?
:backhand_index_pointing_right: What plans do you have for expanding beyond basic lending/borrowing functions, such as integrating yield farming or multi-collateral support?

3 Likes

Thanks for the great questions! @priyankg3

Q1: Is VELT AI building its own lending protocol or integrating existing ones like Aave or Compound?

For our HyperHack MVP, our goal with VELT AI is to build a lightweight, native lending protocol directly on Hyperion. This would allow us to fully explore how natural language interactions, real-time feedback, and intent-driven transaction flows—powered by Alith—can reshape the lending experience from the ground up.

Rather than integrating existing protocols, we’re exploring custom smart contracts to take advantage of Hyperion’s fast finality and parallel execution, which align well with our goal of a responsive, AI-assisted lending flow.

That said, we’re designing the architecture to be modular, so integrations with established protocols like Aave or Compound could be possible in the future—especially if it helps expand liquidity access or drive wider adoption. But for the MVP, our focus is on creating a native-first experience that’s optimized for Hyperion and Alith.

Liquidity Approach (MVP Phase):

For the MVP demo, we plan to simulate liquidity using testnet faucet-driven pools, allowing users to interact with the system in a low-risk environment.

As we move toward mainnet readiness, we’re considering:
• Early-stage incentives to attract initial lenders
• A progressive staking model
• Exploring ecosystem support or grants from Hyperion-aligned partners

This phased approach gives us control over user experience and risk in the early stages, while setting the stage for future expansion and integrations.

Q2: Can users customize loan terms like duration or rate type (fixed/variable)?

For MVP, loan terms will be standardized for simplicity and smooth Alith-driven interactions. Custom loan configurations will be introduced later as we scale user needs and UX complexity.

Q3: How is the credit score calculated?

VELT AI will calculate on-chain credit scores based on:

• Wallet behavior (e.g. past loans, repayment history)

• DeFi activity across supported protocols

• Account age, on-chain reputation, and interactions

We’ll use indexers or analytics tools to fetch this data, and then calculate score logic.

2 Likes

Thanks for the questions Han! Here are some quick answers:

  1. On-chain credit scoring & privacy:

We’ll evaluate wallet activity like repayment behavior and protocol usage using public blockchain data — no off-chain identity or KYC involved. This keeps everything transparent and privacy-respecting.

  1. Safety in conversational actions:

Every action will require explicit confirmation before execution. No transaction will happen without user review and wallet signature, even after an intent is detected.

  1. Risk alerts customization:

Users will be able to opt in/out of alerts and choose how to receive them.

  1. Scalability for multiple users:

The backend will be built to handle many users in parallel, supported by Hyperion’s fast finality and parallel execution. We’ll also scale chatbot sessions with load-aware routing.

  1. Beyond lending/borrowing:

We plan to add yield farming, support for more collateral types, and eventually cross-chain functionality as we grow.

3 Likes

Just straight-forward opinion about the need for this idea:
If i need to borrow money, i am little skeptical to just chat with AI to borrow and lend my real money.
If DEX is complicated, I use CEX.

3 Likes

Thank you for the clear and detailed answers! It’s exciting to see how VELT AI is building a native lending experience optimized for Hyperion.

Just a quick question: Do you plan to support cross-chain lending or will it be Hyperion-native only for now?

2 Likes

Thanks for the detailed answers! I have two quick questions:

1- How do you plan to handle false positives or inaccurate risk alerts to avoid overwhelming users with unnecessary notifications?

2- What measures will you take to ensure seamless user experience as the number of simultaneous chatbot sessions grows?

2 Likes

Thanks for sharing your thoughts Django, I agree, chatting with Ai to borrow or lend real money can feel sketchy, especially if you’re not sure of what’s going on under the hood and I have to clarify that while VELT does explore chat based interactions, we’ll definitely have a proper user interface for anyone who prefers a more traditional experience. You’ll be able to use the dapp like any other.. no need to chat if that’s not your style..

3 Likes

But let’s be honest .. a lot of today’s mainstream tech started off looking weird or untrustworthy…

Think about Uber: getting in a stranger’s car used to be considered dangerous. Now it’s normal.

Or Airbnb: paying to sleep in someone’s house? Sketchy at first. Now it’s a travel essential.

DeFi and AI in finance are going through that same phase and the value lies in transparency, decentralization, and access — especially for those without traditional banking privileges.

CEXes (like Binance, Coinbase, etc.) don’t give loans directly unless they offer centralized margin accounts or collateralized loans.. but you’re trusting a middleman to hold your keys and your funds. If something goes wrong (FTX, Celsius, BlockFi…), users get burned…In contrast, DeFi lending protocols (like Aave, Compound) work with smart contracts and price feeds, so you always know what’s happening with your collateral, your interest rate, and your risk — and no one can rug you behind closed doors..

3 Likes

For now VELT will be native to Hyperion

2 Likes

I appreciate your questions @han — they show that you’re already thinking about user experience and scalability. We understand the risks of receiving too many warnings or alerts, which can lead users to ignore the important ones. To handle that, we plan to allow users to filter the types of alerts they receive. Alerts should be meaningful, not distracting or annoying, and we’ll keep refining the system based on user feedback.

On the chatbot side, we definitely want to make sure the experience stays smooth as more people start using it. We’ll be working closely with our dev team to ensure the system can handle growth without lag or delays, and we’ll keep testing and improving as usage increases. The focus is really on keeping things fast, simple, and reliable for everyone.

3 Likes

I have few questions to get more depth idea of your project.

The proposal states Alith SDK inputs are processed by a “middleware layer that maps them to verified smart contract calls via ABIs and pre-defined schemas.”

  1. How does this middleware technically perform the mapping from potentially varied natural language to precise smart contract function calls and parameters? What mechanisms are in place to manage and update these mappings if smart contract ABIs or the intended schemas evolve?

  2. Beyond user wallet confirmation, what specific security measures are implemented within the pipeline (from Alith SDK input through the middleware to Hyperion) to prevent malicious or crafted language inputs from exploiting the intent parsing mechanism to suggest unintended smart contract parameters or interactions, even if the user is prompted to confirm?

Thank you

2 Likes

Great approach! Customizable alerts and a focus on scalability will definitely help maintain a smooth, user-friendly experience as the platform grows.

1 Like