Builder Spotlight: YieldSense

Builder Spotlight: How YieldSense Brings Hardware-Verified Execution to DeFi on Base

When a DeFi protocol wants to automate something, whether it’s compounding rewards, rebalancing a grid, or triggering a stop-loss, the default pattern is almost always the same: a keeper bot running on a centralized server, a single hot wallet signing every transaction, and a config file somewhere holding the strategy parameters in plaintext.

 

YieldSense, built by Malik Huzaifa, removes that single point of failure entirely. Built on Base, and using Acurast’s decentralized network, every decision, every signature, every strategy parameter now lives inside an Acurast Trusted Execution Environment. The on-chain contract on Base only accepts transactions signed by an attested worker.

 

This post is a short walkthrough of what was built, how Acurast is integrated, and what it tells us about where DeFi automation is heading.

 

Project At A Glance

 

01 / The Centralized Keeper Problem

Four trust assumptions DeFi automation has quietly accepted.

Most DeFi automation today inherits the same four weaknesses. None of them are fundamental. They’re just inheritances from a time when there wasn’t a good way to run off-chain logic with on-chain integrity.

 

02 / How Acurast TEEs Change the Model

Decision logic and signing, both locked in an enclave.

YieldSense uses the Acurast TEE network as the entire decision and signing layer. The user signs an EIP-712 intent in the dashboard, setting their confidential strategy parameters: stop-loss, slippage, optimization range, heartbeat. The encrypted payload is sent to an Acurast TEE worker.

 

Inside the enclave, the strategy engine fetches APR, gas, and price data directly from Base RPCs, runs the decision logic, and only proceeds if expected rewards cover more than 150% of the gas cost. If the trade clears that threshold, the TEE signs the transaction using a hardware-bound P-256 key that never leaves the enclave.

 

The YieldSenseKeeper contract on Base then verifies the hardware attestation before triggering execution. Anything not signed by a whitelisted Acurast worker is rejected outright.

 

The net effect is straightforward: strategy parameters never leave the enclave, signing keys never touch a server, and the on-chain contract has cryptographic proof of where every transaction originated. That eliminates the single point of failure, keeps the strategy confidential, and makes every step of execution verifiable on-chain.

The three-plane architecture. User intent flows top to bottom, through the enclave, to attested on-chain settlement.

 

03 / What’s Live Today

An Aerodrome USDC/AERO autocompounder running the full TEE pipeline.

 

The Mainnet MVP runs an Aerodrome USDC/AERO autocompounder on Base, fully driven by the Acurast TEE pipeline. Real APR data ingestion, real gas validation, real harvests, real attested signatures. The dashboard surfaces it all: net positions, projected alpha, gas optimization, and a live Guardian Ledger of every TEE action.

 

Underneath, every cycle the keeper runs through the same five-phase sequence. None of it advances unless the previous phase produces a hardware-attested result.

The harvest cycle. Each step is hardware-attested before it advances.

Every action lands in the Guardian Ledger as it happens, with the hardware address that signed it.

 

 

This is what hardware attestation looks like in real-time: every execution and attestation tagged, timestamped, and tied to a specific Acurast processor address.

 

Two contracts handle the on-chain side:

 

04 / A Verifiable Foundation for Grid Trading

Hardware-attested grid signals, ahead of live execution.

The grid-trading side of YieldSense is intentionally cautious, and worth understanding precisely.

 

Right now, the Grid Keeper operates as a price-triggered signed executor. The TEE monitors a configured pool (currently AERO/USDC), compares the live price against fixed reference levels defined in GRID_CONFIG_JSON, and when a price moves beyond a set triggerPercent, the enclave calculates the theoretical PnL delta for that grid level and signs a cryptographic proof of the action. The processor calls executeTrade() on the keeper contract, and the contract records the trade as a verifiable event on-chain.

 

What it does not do yet: interact with DEX routers, manage token inventory, or shift balances between users. The grid logic produces hardware-attested audit logs, not live position changes.

 

This is by design. Before YieldSense moves real inventory, the team wanted to prove the enclave can correctly identify, evaluate, and sign trade triggers, with every action recorded on-chain as a hardware-attested event. The Guardian Ledger gives auditors a complete record of what the agent saw, what it decided, and what it signed. Live execution is the natural next step from the same foundation.

 

05 / What This Tells Us About Acurast for DeFi

A pattern that generalises beyond yield harvesting.

YieldSense is doing something more than building a single yield product. It’s showing that the trust assumptions DeFi automation has been carrying for years aren’t structural, they’re situational.

 

The same pattern YieldSense uses for compounding and grid signals generalises directly to other primitives: AI agents managing capital under provable constraints, cross-chain settlement with attested off-chain logic, oracle attestation with hardware-backed signing, conditional payments triggered by enclave-verified events. None of these need a new framework. They need the keeper layer to be trust-minimized. That’s the layer Acurast TEEs provide.

 

About the Builder

YieldSense is built by Malik Huzaifa, a software architect and blockchain engineer focused on trust-minimized autonomous systems. The project is the output of his research into TEEs and their application to decentralizing complex DeFi strategies. He’s active on X and LinkedIn.

 

This is the network YieldSense plugs into. Every figure above updates as the network grows.