Aqualis Protocol - Dual Yielding in DeFi

Hyperion HyperHack Ideathon
1

Project Name

Aqualis Protocol

Problem Statement

Currently, DeFi protocols are fragmented in liquidity and cannot communicate with each other between lending and trading, where billions of dollars of liquidity is potentially going to waste. Aqualis has created a asset multi-utilization (AMU) algorithm, but it is currently implemented as a backward looking algorithm that can’t dynamically adjust to live data. We aim to use AI to create a forward looking algorithm that can use current market data and sentiment to adjust the liquidity needs of the trading/lending pools.

Solution Overview

Aqualis Protocol wants to build an on-chain AI algorithm that can look at on chain metrics such as price, volume and even potentially off-chain metrics like sentiment using oracles to create an algorithm that can predict the liquidity needs of DEXs and lending. This aims to optimize the liquidity provided to Aqualis Protocol to maximize capital efficiency, delivering lower fees and higher yield for depositors.

This will work in a decentralized way using an on-chain algorithm, similar to how the current AMU has been implemented but using AI instead, different from current rehypothecation platforms that exist which optimize yield on third party protocols while having centralized control of funds.

Project Description

  • An on-chain AI algorithm using market data to optimize yield. The algorithm will take metrics like volume, price, trends, social sentiment and possibly more factors into account to try and predict how much liquidity can be safely used for lending, and when it needs to be brought back into the DEX side.
  • The exact technological implementation and tech stack is to be decided
  • Users can simply deposit liquidity onto Aqualis and watch the magic happen!
  • If this works as intended, this could bring forth a new level of capital efficiency in DeFi, allowing DEXs to not only compete with CEXs who traditionally have much lower fees, but also compete with TradFi as more RWAs like equities, derivatives and commodities are brought on chain.

Community Engagement Features

Users will be able to claim tokens on a testnet faucet every day, users can earn points proportional to their activity for a certain feature. This encourages users to test all features we need tested to optimize their points.

  • Adding liquidity: 1000 points per day spread between all users
  • Removing liquidity: 0 points
  • Trading: 1000 points per day spread between all users
  • Adding collateral: 0 points
  • Removing collateral: 0 points
  • Opening a loan: 1000 points per day spread between all users (based on open loans)
  • Repaying a loan: 0 points
  • Playing with on chain metrics to alter the AMU AI: 0 points, done by the team to test the effectiveness of the AMU AI

By gamifying this by creating set pools of points per feature, users are encouraged to try all the features. If we add rewards for the points, users are encouraged to think ahead to optimize their points. Since the faucet distribute a set amount per day, users can also think about how to optimize their funds, do they want to farm points early on or try and earn testnet tokens through yield.

Getting Involved

Any users who have experience in AI is welcome to give input as we are relatively new to this. Testers are always welcome too!

6 Likes
2

Thank you so much! I have a question that’s been on my mind.
How will Aqualis Protocol’s on-chain AI algorithm balance real-time market data and sentiment to dynamically optimize liquidity allocation between DEX lending and trading pools?

1 Like
3

Hello,

Can you walk through a real-world example of how the AI would shift liquidity between DEX and lending markets—say, during a high-volatility event like a CPI release or BTC flash crash?

If the AI over-allocates liquidity to lending and there’s a sudden loan default or market downturn, how is the DEX side protected from becoming illiquid?

4 Likes
4

Hi Han, thanks for your interest in the submission.

For some context, the more DEX liquidity we can allocate into lending the more efficient capital can be. However, the risk is running out of liquidity for a certain asset or lending interest rates becoming too variable.

With the use of AI, we believe there is an opportunity to better manage these tradeoffs to create an ecosystem that can provide both good yields for depositors, and low fees plus great usability for traders and borrowers. How AI can help with this is by using real time (or close to real time) data to estimate the liquidity needs for the DEX better than we can do as forward looking humans by taking into account large amounts of aforementioned data. The current AMU algorithm is static, which means it changes only when the team (later the DAO) decides it needs to be changed.

2 Likes
5

Hi, happy to walk through a real-world scenario.

Let’s take a high-volatility event like a CPI release. In the lead-up to such events, our AI trained on both on-chain data and off-chain signals like macroeconomic calendars may anticipate increased trading activity and volatility. In response, it could reallocate liquidity from lending back into the DEX to ensure adequate liquidity.

While this may seem intuitive, our goal is for the AI to outperform simple rule-based or human-driven rebalancing by taking into account a broader set of variables (e.g., funding rates, historical behavior under similar events, sentiment analysis) and adapting in real time.

Now, on to your second question: If the AI over-allocates liquidity to lending and a sharp downturn or loan default occurs, how is the DEX side protected?

First, it’s important to note that the stablecoins in our DEX aren’t gone, they’re simply lent out in an overcollateralized manner but access to that liquidity is temporarily unavailable. It’s worth noting that in the event of a downturn, borrowers who have used volatile assets as collateral may be liquidated, returning stablecoins to the protocol. This naturally replenishes liquidity on the DEX side.

However, if liquidations are insufficient or delayed, and the DEX experiences a short-term stablecoin crunch, we have a contingency in place. The protocol can access emergency liquidity by borrowing from external lending markets such as Aave, using the surplus crypto assets flowing into the DEX (as users sell volatile assets) as collateral.

Key considerations:

  • Repayment: Since our stablecoins are still present and just lent out and collateralized, we are not over-leveraging. The loan is simply a temporary bridge while funds are in use.
  • Collateral: The inflow of crypto assets during a downturn provides ample collateral to borrow stables.
  • Risk of loss from borrowing: If all Aqualis stablecoins are lent out, interest rates will spike due to the utilization curve. This high return on deposits makes it highly unlikely that external borrowing at lower rates would be unprofitable. In fact, interest rate arbitrage will likely attract external capital, quickly restoring equilibrium.

This layered approach AI-driven allocation, automated liquidations, and emergency bridging, ensures that Aqualis remains resilient, adaptive, and liquid even in turbulent market conditions.

The risks of a “black swan event” where blue chips crash substantially within seconds is the only real risk of this set up, but it’s worth noting all DeFi lending operations carry the same risks of bad debt in this scenario, it’s just a matter of how much bad debt which depends on the liquidation thresholds.

3 Likes
6

Hey @cryptoeater , Couple of questions that I have in mind regarding your project that might be crucial.

  1. Why does DeFi need an AI model to determine liquidity needs when simple heuristics or rebalancing curves often suffice?
  2. What does it mean to have “on-chain AI”? AI models are typically large and compute-heavy. How would this be executed and updated on-chain?
  3. If you’re relatively new to AI, why would users trust that the algorithm is making sound predictions about millions in liquidity?
  4. Is optimizing internal Aqualis liquidity usage even a top-of-mind problem for most LPs, or is it yield and security?
2 Likes
7

Thanks for the context. Makes perfect sense. The shift from a static AMU to an AI-driven, adaptive model sounds promising. Curious though: how do you plan to balance responsiveness with stability, especially during sudden market shifts?

8

Thanks for the reply.

Will definitely be keeping an eye on how Aqualis evolves. :slight_smile:

2 Likes
9

Thanks for the context

2 Likes