Inspiration
Money moves between businesses every day, yet payment decisions are usually made in isolation.
A business may have enough cash to pay one supplier but not another. That supplier might be waiting on the payment to cover its own obligations. A single timing decision can affect an entire network of businesses.
The problem isn't always a lack of money. Sometimes it's where the money is, when it becomes available, and whether businesses can agree on better payment terms.
We built Elide to address that problem.
What it does
Elide is an AI-powered liquidity optimization and payment coordination platform built around PayPal.
It analyzes participating businesses' invoices, cash reserves, payment deadlines, available liquidity, and financial policies to identify opportunities for smarter payment coordination.
Rather than simply recommending which invoice to pay first, Elide considers how payment timing affects connected businesses.
Its core capabilities include:
- Liquidity intelligence: Identifies bottlenecks, protected reserves, funding shortfalls, and feasible payment sequences.
- Network optimization: Mathematically evaluates payment schedules against alternative strategies, considering timing, fees, discounts, and financial constraints.
- AI-assisted negotiation: Interprets merchant policies, drafts bounded counteroffers, and explains opportunities using a real language model.
- Independent approvals: Both parties must agree to revised terms, and the payer separately authorizes any payment.
- PayPal execution: Creates approved invoices, directs buyers to PayPal's hosted checkout, and verifies the resulting transactions.
- Financial reconciliation: Records actual fees, currency conversion, settlement outcomes, and duplicate-resistant accounting entries.
Crucially, AI never controls the money. Financial constraints are enforced by deterministic code, and merchants retain authority over their decisions.
How we built it
Elide combines financial optimization, AI, and payment infrastructure.
The frontend uses React, TypeScript, and Tailwind CSS. The backend uses Python, FastAPI, and PostgreSQL, with durable transaction records, merchant-scoped authorization, and an append-only accounting journal.
Our optimization engine searches for feasible payment sequences while enforcing cash reserves, due dates, fees, commitments, and merchant consent. A separately implemented verifier checks the results.
The AI layer uses Qwen3-235B-A22B-Instruct through Nebius Token Factory for financial policy interpretation, negotiation drafts, and explanations. Every model output passes deterministic validation before it can affect a workflow.
Payments use the PayPal Sandbox Invoicing API and PayPal-hosted checkout. Provider records are fetched independently and reconciled against Elide's financial ledger.
Accomplishments
We successfully demonstrated an end-to-end payment through PayPal Sandbox.
Two test merchants agreed on a $1 early-payment discount for a $25 invoice. Elide verified the financial conditions, collected independent approvals, and created and sent the resulting $24 invoice through PayPal.
The buyer completed the hosted Sandbox payment. PayPal independently confirmed the transaction, including a $1 processing fee and $23 net proceeds converted into CAD.
Elide reconciled the payment, balanced the accounting records, and replayed the reconciliation without creating duplicate entries.
We also evaluated our optimizer against independent verification and simpler payment strategies, and tested our AI policy system using real model inference and synthetic financial scenarios.
The potential at scale
Our live PayPal Sandbox demonstration deliberately used a small $25 invoice and a $1 discount. These amounts were chosen to validate the entire financial workflow safely, not to represent the scale of the problem Elide is designed to address.
Businesses routinely deal with much larger obligations, overlapping payment deadlines, and multiple suppliers competing for limited liquidity.
For perspective, a negotiated 1% discount on a $100,000 invoice represents $1,000 in gross savings to the buyer before financing costs and other expenses. Whether that arrangement benefits both parties depends on the supplier's value of earlier payment, processing fees, reserves, and other financial constraints.
The larger opportunity goes beyond individual discounts. Coordinating payment timing across a network could help some businesses meet obligations that otherwise would be delayed, without introducing new borrowing or creating artificial liquidity.
Elide is designed to find those opportunities, verify that they are financially feasible, and coordinate agreements that participating merchants are willing to accept.
The Sandbox transaction proves the payment and reconciliation mechanism works. The economic benefits of operating at a larger scale remain to be validated with real merchant data.
Challenges we ran into
The hardest challenge was ensuring financial correctness.
A successful payment doesn't necessarily mean its proceeds are immediately spendable. Fees, currency conversion, reserves, delayed availability, and competing obligations can break seemingly feasible payment chains.
We also learned that optimizing a network doesn't automatically produce better results than simple strategies. Elide explicitly measures incremental improvement rather than treating gross payment volume as newly created liquidity.
Another challenge was separating AI recommendations, merchant consent, payment authorization, and actual provider-confirmed settlement.
What's next
Our long-term vision is to make Elide a financial coordination layer that businesses can use alongside PayPal.
Next, we're expanding payment lifecycle coverage, improving optimization and liquidity modeling, strengthening merchant connectivity, and pursuing real-world validation with consenting businesses.
Our current evidence comes from controlled PayPal Sandbox transactions, real AI inference, and synthetic financial networks. Real-world economic improvement remains a hypothesis to validate.
Elide's goal is simple: help businesses make better use of the liquidity they already have, while keeping every payment accountable, permissioned, and verifiable.
Remaining development
Elide is an actively developed platform. The core financial engine, AI-assisted policy workflows, merchant application, independent approvals, and end-to-end PayPal Sandbox integration have been implemented and tested.
The remaining work focuses on taking that foundation further:
- Financial intelligence: Refine discount economics, support more flexible payment scheduling, and improve liquidity availability modeling.
- Payment lifecycle: Expand webhook verification, refunds, reversals, exception recovery, and settlement auditability.
- Merchant connectivity: Introduce broader permissioned account integrations and more reliable financial evidence beyond merchant attestations.
- Optimization: Scale beyond small synthetic networks, improve incentive alignment, and evaluate performance against stronger financial benchmarks.
- AI and machine learning: Expand proactive recommendations and investigate forecasting models using appropriately validated financial data.
- Production readiness: Strengthen authentication, deployment, monitoring, security, accessibility, and operational recovery.
- Real-world validation: Test with consenting businesses and measure actual payment timing improvements, economic benefits, and PayPal's incremental value.
We intend to continue developing Elide beyond the hackathon, with financial correctness, merchant control, and measurable business value remaining central to the architecture.
Demonstration video
Because of the submission time constraints, we were unable to prepare the complete narrated product walkthrough we originally planned. A more comprehensive demonstration showing Elide's merchant application, liquidity optimization, AI workflows, PayPal payment execution, and reconciliation will be published shortly.
In the meantime, the public repository contains the application source, technical documentation, testing evidence, and results from the completed PayPal Sandbox integration.
Built With
- ai
- docker
- fastapi
- nebius
- optimization
- paypal
- playwright
- postgresql
- pytest
- python
- qwen
- react
- sqlalchemy
- tailwind-css
- typescript
