Inspiration
The global transition to a sustainable economy is currently stalled by a fundamental paradox: Transparency vs. Privacy.
On one side, regulators and banks demand granular data to verify ESG claims and avoid the massive reputational and legal risks of "Greenwashing" (with fines averaging $2M+ per incident). On the other side, suppliers refuse to share raw production data—their electricity usage, material sourcing, and logistics—because these are trade secrets.
We realized that current solutions suffer from the "Garbage In, Verified Out" problem. Most blockchain projects verify the model, but they don't verify the data source. We were inspired to build SustainSwift to bridge this gap. We wanted to move sustainability from a passive, annual PDF report into a real-time, verifiable financial asset that is usable for small suppliers.
Further Thoughts: The Cost of Inaction CFOs and Sustainability Officers face a critical gap between the speed of business and the speed of reporting. Data Silos: Financial reports, supply chain audits, and CDP disclosures live in disconnected systems. Compliance Liability: With tightening SEC and CSRD mandates, delayed or inaccurate reporting is a legal liability carrying potential fines averaging $2M+ per incident [1]. Competitive Disadvantage: 72% of investors now penalize companies with opaque ESG reporting [2].
What it does
The transition to a sustainable economy is stalled by a fundamental paradox: Transparency vs. Privacy. Regulators and investors demand granular emissions data to verify ESG claims ("Greenwashing" risk), but suppliers refuse to share raw production data due to trade secrets ("Competitive" risk). Furthermore, current solutions suffer from the "Garbage In, Verified Out" problem—verifying the model without verifying the data source. This project introduces SustainSwift, a decentralized AI agent protocol that establishes a "Triple Truth" architecture: Source Truth: Using zktls to prove data origin.
Compute Truth: Running GenOptima models inside AWS Nitro Enclaves.
Settlement Truth: Using Caffeine AI canisters for real-time Dynamic Sustainability-Linked Loans (SLLs).
By transforming ESG from a passive report into a verifiable asset, SustainSwift enables banks to adjust interest rates dynamically based on mathematically proven sustainability performance.
How we built it
Introduction video to Zero-Knowledge-Proof: https://www.youtube.com/watch?v=lUTv9NHkuR4
Technical Architecture SustainSwift replaces the "Trust me" model with a cryptographic "Triple Trust" layer.
A)The Trust Layer Components
Source Truth: The Digital Notary (zktls) Problem: How do we know the input data isn't a fake CSV? Solution: We utilize zktls (Zero-Knowledge TLS) to generate a session proof directly from the utility provider's API (e.g., api.duke-energy.com). Mechanism: Proof_source = Verify(HTTPS_session, UtilityProviderKey) Result: Proof that data originated from a trusted source without revealing login credentials or raw bill details.
Compute Truth: The Isolated Brain (AWS Nitro + GenOptima) Problem: How do we know the supplier ran the correct model? Solution: The GenOptima AI model runs inside an AWS Nitro Enclave. This isolated execution environment provides a cryptographic attestation that the code and model weights were not tampered with during execution. Feature: Counterfactual Analysis. Unlike standard reporting tools that only look backward, GenOptima runs "what-if" simulations inside the enclave. A supplier can ask, "What if I switch to 80% Solar next month?" The system simulates the new score and issues a "Provisional Certificate." This certificate serves as a verifiable promise of future performance, allowing suppliers to secure "Green Transition Loans" to fund the upgrade before the capital expenditure is made.
Verification Truth: The Settlement Layer (Caffeine AI) Problem: Where does the final verdict live? Solution: A Caffeine AI Canister acts as the settlement layer. It does not store raw data. Instead, it stores the Merkle Root of the validated proofs. Function: The canister verifies the zktls signature and the AWS Nitro attestation. If valid, it triggers the smart contract to adjust the loan terms.
B)System Workflow The Cryptographic Flow: Step 1 (Source Truth): Supplier Agent connects to Utility API via HTTPS. zktls generates a Session Proof (hiding credentials). Step 2 (Compute Truth): AWS Nitro Enclave receives the data + session proof. It runs the GenOptima model and generates a zk-Attestation. Step 3 (Verification): The Proof Payload (zktls Sig + Nitro Attestation) is submitted to the Caffeine AI Canister. Step 4 (Settlement): Caffeine AI validates the proof and updates the State Merkle Root. The Bank's Smart Contract queries this root to trigger a Dynamic Rate Drop.
C) Methodology & Novelty
1 The Verifiable Scoring Function
Unlike opaque "black-box" ESG ratings, SustainSwift calculates the Sustainability Score (S) as a deterministic function of proven inputs. This formal definition ensures that the score is mathematically reproducible: S = [ Σ (w_i · I_i) · α_sector ] / Revenue Where: w_i represents the emission factors verified by the GenOptima model. I_i is the input verified by zktls. α_sector (Alpha) is the sector-specific weighting constant. Governance Note: Crucially, Alpha is not hardcoded or centralized. It is a governance parameter set by the lending consortium or a sustainability DAO. This prevents centralization and allows the protocol to dynamically adjust risk weights for "dirty" industries (e.g., Cement vs. Software) without requiring core code updates.
2 Solving the "Last Mile" Problem
Traditional blockchain solutions fail at the "Oracle" layer—getting real-world data on-chain trustlessly. By combining zktls (Source) with AWS Nitro (Compute), we close the loop, creating the first end-to-end trustless ESG pipeline.
Challenges we ran into
The "Oracle" Problem: The hardest part of blockchain ESG is getting real-world data on-chain trustlessly. Building the "Source Truth" layer using zktls was complex, as we had to ensure the proof bound the data to a specific domain certificate without leaking the session key.
Privacy vs. Verifiability: Balancing the need for the bank to trust the score while ensuring the supplier's raw data never left the AWS Enclave required strict architecture design. We had to ensure the GenOptima model could run "Counterfactual Analysis" (what-if scenarios) inside the enclave without exposing the weights to the user.
The "Corrupt Source" Risk: We realized that even with zktls, a corrupt utility insider could falsify records. We had to design a defense-in-depth strategy involving multi-source corroboration and statistical anomaly detection to flag outliers.
Accomplishments that we're proud of
The "Green Transition Loan": We enabled a feature where suppliers can run simulations ("What if I switch to solar?") to get a Provisional Certificate. This allows them to secure funding before they make the capital expenditure.
End-to-End Trustless Pipeline: We believe we have created the first ESG pipeline that closes the loop between the physical world (utility APIs) and the financial world (Smart Contracts) without a trusted middleman.
Automated ROI: We demonstrated that this system can reduce costs while reducing the time-to-verification and Risk.
What we learned
Compliance is a floor, not a ceiling: We learned that banks aren't just looking to avoid fines; they are looking for "Alpha." They want to capture the value of low-risk, high-sustainability borrowers.Governance is key: Hardcoding risk weights is dangerous. We learned the importance of the $\alpha_{sector}$ variable—a governance parameter that allows the protocol to adjust risk weights for different industries (e.g., Cement vs. Software) without rewriting the core code.
What's next for A PrivacyPreserving Protocol for Verifiable ESG Intelligence
Honestly, our immediate goal is simple: We hope to win the 2025 Sustainable Finance Hackathon.
Beyond that, we didn't build SustainSwift to launch a startup; we built it to prove a point. We believe the "Transparency vs. Privacy" paradox is the single biggest blocker to a sustainable economy, and we wanted to show that technology can solve it.
We hope this protocol serves as a blueprint for the industry. If this architecture inspires even one bank or one major supplier to rethink how they verify ESG data—moving from "trust me" to "verify me"—we will consider it a success.
We want to see a future where "Greenwashing" is mathematically impossible because the data speaks for itself.
Log in or sign up for Devpost to join the conversation.