Build peer-to-pool lending markets that protect liquidity under extreme volatility.

A lending protocol is not an ERC-20 contract with a borrow button. It is an algorithmic risk and solvency machine.

Under normal market conditions, almost any lending contract functions. But when asset prices drop 40% in two blocks, protocols fail because of three specific engineering oversights: oracle latency allowing underwater positions to become unliquidatable bad debt, flash loan griefing draining reserve pools, and rigid interest rate models causing liquidity crunches where depositors cannot withdraw. PROPELOO builds institutional lending architecture: isolated lending markets that prevent single-collateral contagion, dual-oracle pricing with fallback circuit breakers, dynamic Dutch auction liquidations, and invariant property testing that formally verifies protocol solvency under extreme liquidation pressure.

The architectural layers of a production DeFi lending protocol.

From interest rate models to automated keeper liquidation engines, every layer is built for mathematical solvency.

System Layers

• Core Pool & Accounting Layer: Interest-bearing receipt tokens (aTokens/cTokens), debt tracking via normalized indices, reserve pools and collateral balance segregation.
• Interest Rate & Utilization Engine: Kinked-curve dynamic interest rate models adjusting borrow APY automatically based on pool utilization to incentivize liquidity rebalancing.
• Oracle & Price Aggregation Layer: Dual Chainlink and Pyth oracle aggregation, time-weighted average price (TWAP) sanity checks, and stale price circuit breakers.
• Liquidation & Keeper Engine: Dutch auction and fixed-discount liquidation mechanics, flash liquidation callbacks, and automated keeper bot infrastructure.
• Flash Loan & Liquidity Vaults: Atomic uncollateralized flash loans with premium fee accrual to protocol reserves and reentrancy-guarded execution hooks.

How we engineer decentralized lending protocols.

Protocol solvency is an immutable law. Every line of code must defend collateral adequacy.

• Bad debt must be isolated, never socialized across markets

  Shared pool designs mean a single volatile or illiquid collateral asset can drain the entire protocol reserves. We build isolated lending pairs and segregated debt pools so high-risk or long-tail assets can never threaten core stablecoin liquidity.

  Axiom: ISOLATED RISK ARCHITECTURE

• Oracles require redundant verification and staleness guards

  Single-source oracles are the primary attack vector in DeFi lending. We implement dual-oracle aggregation with deviation bounds, heartbeat checks, and circuit breakers that pause liquidations if price feeds disagree.

  Axiom: DEFENSE-IN-DEPTH ORACLES

• Liquidation speed must exceed market price decline velocity

  If liquidators cannot profitably liquidate an unhealthy loan before collateral value drops below borrowed debt, the protocol absorbs permanent insolvency. We design Dutch auction liquidations and low-gas keeper execution contracts.

  Axiom: RAPID SOLVENCY PRESERVATION

Architectural decisions in DeFi lending engineering.

These choices define protocol solvency, capital efficiency and vulnerability profile.

• Shared pool-to-peer (Aave v3) vs isolated lending pairs (Morpho/Euler)?

  Impact: Hybrid modular architecture. Core liquid assets (ETH, WBTC, USDC) share a high-efficiency pool, while governance tokens, yield-bearing assets, and RWAs reside in isolated markets with strict risk ceilings.

  • Shared pool — maximum capital efficiency and liquidity depth, shared systemic risk across all accepted collateral
  • Isolated pairs — zero contagion between assets, lower capital efficiency for cross-margin traders
  • Hybrid modular pools — core tier-1 assets in shared pool, long-tail and RWA assets in isolated markets

• Interest rate curve model: two-slope kinked curve vs PID controller?

  Impact: Two-slope piecewise linear kinked model with governance-configurable kink point (80-85%). Provides transparent predictability for borrowers while providing exponential rate surges to protect withdrawer liquidity during crunches.

  • Two-slope kinked curve — predictable utilization target (e.g. 80%), steep slope beyond kink to penalize full pool exhaustion
  • PID automated rate controller — adapts rates dynamically to market benchmark rates, complex on-chain tuning

• Liquidation mechanism: fixed bonus vs Dutch auction?

  Impact: Dynamic Dutch auction liquidation with floor cap. Ensures liquidators are incentivized without needlessly stripping borrower equity during temporary liquidity dislocations.

  • Fixed discount bonus (5-10%) — simple to execute, susceptible to MEV gas wars and over-penalization in volatile dumps
  • Dutch auction liquidation — discount starts low and increases over time, captures optimal price for borrower and protocol

• Oracle architecture: primary oracle with TWAP fallback vs multi-source medianizer?

  Impact: Dual-feed medianizer (Chainlink + Pyth) with 1.5% maximum allowable price deviation threshold and automatic fallback pause if either feed reports stale data.

  • Chainlink primary with Uniswap V3 TWAP secondary fallback — robust for high-cap assets
  • Chainlink + Pyth medianizer with deviation circuit breaker — sub-second updates for multi-chain and L2 deployments

DeFi lending architectures PROPELOO builds.

• Institutional Overcollateralized Lending

  Enterprise lending markets for compliant institutional participants with whitelisted collateral assets, KYC integration, and automated reporting.

• RWA & Private Credit Markets

  Lending protocols accepting tokenized real-world assets, private credit notes, and trade finance receivables as loan collateral.

• Liquid Staking & Restaking Lending

  Leveraged looping protocols supporting LSTs (stETH) and LRTs with automated debt rebalancing and yield-maximization strategies.

• Flash Loan Protocol Vaults

  Dedicated uncollateralized flash loan reserve contracts earning fee revenue for liquidity providers from arbitrageurs and liquidators.

• Permissionless Lending Factory

  Factory contract deployment enabling users to spin up isolated lending markets for any arbitrary ERC-20 token pair with custom oracles.

The DeFi lending engineering stack.

Battle-tested Solidity smart contracts and high-frequency keeper infrastructure.

• Smart Contracts

  Stack: Solidity 0.8.24, Foundry, OpenZeppelin v5, Hardhat, Huff (gas-optimized math)

• Oracles

  Stack: Chainlink Price Feeds, Pyth Network, Uniswap V3 TWAP, Chronicle Oracles, Chainlink PoR

• Security & Verification

  Stack: Slither, Echidna Invariant Fuzzing, Certora Prover, Halmos Symbolic Testing

• Keeper & Liquidation Bots

  Stack: Rust, Go, Flashbots Protect, Tenderly Virtual Testnets, Subgraphs

• Frontend & Analytics

  Stack: Next.js 14, viem / wagmi, The Graph Subgraphs, Dune Analytics API, PostgreSQL

Solvency and smart contract security in lending protocols.

Lending protocols are the highest-value targets in Web3. Security must be proven mathematically.

• Reentrancy in token callbacks

  ERC-777, ERC-1155 and certain ERC-20 tokens feature transfer callbacks that can reenter deposit or borrow routines. NonReentrant guards on all external state transitions and CEI pattern enforcement.

• Oracle manipulation and stale pricing

  Flash loan attacks inflating AMM pool reserves can corrupt spot price oracles. Strict reliance on decentralized off-chain oracles (Chainlink/Pyth) with minimum heartbeat verification and deviation ceilings.

• Bad debt accumulation

  If collateral drops faster than liquidators can clear positions, the protocol becomes insolvent. Reserve backstop funds and automated protocol debt auctioning mechanics to clear underwater accounts.

• Rounding errors in interest calculation

  Compounding per-second interest math can accumulate fractional truncation errors. Math library implementations using ray/wad precision (10^27 and 10^18 fixed point) with rounding in favor of protocol solvency.

• Flash loan callback exploitation

  Malicious flash loan recipients attempting state manipulation. Flash loan logic requires strict validation that repaid principal plus fee is returned before execution completes.

From economic model to mainnet deployment.

1. 01. Economic Modelling

   Simulation of interest rate curves, LTV thresholds, and liquidation incentives under historical market stress test scenarios.

2. 02. Contract Architecture

   Core ledger, debt tokens, collateral vaults, interest rate calculators, and oracle adapter development in Foundry.

3. 03. Invariant Fuzz Testing

   Comprehensive property-based fuzz campaigns with Echidna: verifying protocol total debt equals active borrow balances.

4. 04. Keeper & Liquidation Bots

   Development of open-source automated liquidator bots and keeper networks to guarantee liquidation execution.

5. 05. Formal Verification & Audit

   Mathematical verification of core solvency invariants and dual independent third-party smart contract audits.

6. 06. Testnet & Staged Launch

   Deploy to Sepolia testnet, simulate high-volatility liquidations with simulated whales, and staged capped mainnet rollout.

DeFi Lending Protocol Engagements

Decentralized money markets, isolated CDP protocols and automated keeper liquidation systems.

• Multi-Asset Decentralized Lending Protocol (M TVL)

  Challenge: DeFi foundation needed high-performance lending protocol supporting cross-margin lending and borrowing across 12 crypto assets on Ethereum L2.

  Architecture: Modular money market protocol with kinked interest rate curves, dual Chainlink/Pyth price oracle adapter, dynamic Dutch auction liquidator, and custom The Graph indexing layer.

  Outcome: Scaled to M peak TVL within 4 months of launch. Successfully liquidated .4M in underwater collateral during a 35% market drop with zero bad debt incurred.

• Isolated Risk Lending Protocol for Yield-Bearing Collateral

  Challenge: Ecosystem protocol required lending markets allowing users to borrow stablecoins against yield-bearing ERC-4626 vault tokens without risking core pool contagion.

  Architecture: Segregated pair architecture where each collateral asset is paired exclusively against USDC. Isolated borrow caps, custom interest rate curves reflecting vault staking yields, and Flashbots-integrated liquidation bots.

  Outcome: Over M in yield-bearing assets deposited. Property-based fuzz tests confirmed 100% solvency across 5,000 simulated market crash iterations.

• Zero-Capital Flash Loan Vault & Arbitrage Network

  Challenge: Trading desk needed flash loan routing contract aggregating uncollateralized borrow liquidity across multiple lending pools with minimal gas overhead.

  Architecture: Ultra-gas-optimized assembly (Yul) flash loan callback dispatcher with atomic multi-DEX routing, automatic slippage guards, and gas refund mechanics.

  Outcome: Processed over M in cumulative flash loan arbitrage and liquidation volume. Protocol earned ,000 in fee revenue distributed directly to liquidity providers.

Frequently Asked Questions