Park Pal

Inspiration

Last summer, we tried to attend a local sheep shearing event at Kelsey Creek Farm. When we arrived, the only parking lot was completely full, and the event ended up getting shut down due to the resulting traffic congestion. However, while we were fighting for a spot, we noticed that the surrounding neighborhoods had hundreds of completely empty driveway spaces.

This isn’t just an isolated incident; in Seattle alone, there are 3.3 unused driveway spaces per household. That represents thousands of dollars in wasted space and pointless traffic. We realized we needed to build a solution to fix this massive inefficiency in urban space utilization.

What it does

ParkPal is the Airbnb for parking. It is a peer-to-peer marketplace that lets homeowners rent out their unused driveways to drivers who desperately need parking.

Because our surveys indicated that 65% of homeowners are worried about safety and property damage from strangers, ParkPal isn't just a booking button—it's a comprehensive trust framework. The platform includes mandatory ID and license verification, a secure photo check-in system, and structured escrows to ensure both drivers and hosts are protected.

How we built it

We split our build into two core phases: Market Intelligence and the Platform Architecture.

Market Intelligence: We built a custom GIS algorithm using Replit. Initially, we only used parking ticket data, but that only showed demand. Our new algorithm layers in supply, demand, walkability, and competition to identify "Goldilocks zones"—areas with both high demand and available residential supply.

Platform Architecture:

Frontend & Compute: We built the application using the Next.js 16 App Router hosted on Vercel. We use a hybrid rendering approach: Client Components manage our interactive Mapbox GL UI, while Server Components handle Server-Side Rendering (SSR) for read-heavy routes.
Data Layer: The backend relies entirely on Amazon DynamoDB (configured for on-demand capacity) and Amazon S3 for private asset storage. We access AWS purely server-side via the AWS SDK.
Security & Auth: We implemented a completely stateless authentication system using jose and bcrypt with JWTs stored in httpOnly cookies.
Payments: Escrow and transactions are handled via Stripe Checkout.

Challenges we ran into

1. Algorithm Accuracy Our initial GIS algorithm mistakenly identified downtown Seattle as a prime spot due to massive demand. However, looking closely at the supply side, there are almost no residential houses there to host drivers. We had to rewrite the model to actively balance supply metrics against demand to find viable launch zones.

2. Shifting to a NoSQL Mindset Scaling a peer-to-peer marketplace on a standard SQL database leads to bottlenecks due to complex JOIN operations. We had to completely rethink our data model to use DynamoDB. By utilizing a table-per-entity design and building Global Secondary Indexes (GSIs) mapped strictly to our access patterns (e.g., by_host, by_listing), we eliminated SQL joins to ensure our query time complexity remains constant regardless of scale:

$$ \lim_{N \to \infty} T_{\text{query}}(N) = O(1) $$

3. Trust & Safety Hurdles We conducted a survey of over 260 people and found a perfect mismatch: 87% of drivers need parking where 77% of homeowners have empty space. The challenge was bridging the trust gap. We had to pivot our development time to build a robust legal and trust framework (ID checks, photo uploads directly to S3 via presigned URLs) to overcome host hesitations.

Accomplishments that we're proud of

Data-Driven Validation: Conducting primary research with 260+ users to mathematically prove our market mismatch and directly guide our product's safety features.
The Alpha Score Algorithm: Successfully layering multiple complex datasets (walkability, ticket data, housing density) to create a proprietary mapping tool that dictates our exact go-to-market strategy.
Enterprise-Grade Architecture: Designing a completely serverless, zero-join, stateless web application that can handle massive traffic spikes without database lockups.

What we learned

We learned that writing code is only half the battle. Our user surveys taught us that a marketplace cannot function purely on convenience; it must be built on trust. Technically, we learned how to architect for massive scale from day one, deeply understanding NoSQL access patterns and the power of separating compute layers from storage networks.

What's next for Park Pal

Our immediate goal is to launch Seattle’s first residential parking network. To do this, we are actively seeking:

Insurance Partnership: Securing a provider to give our hosts professional coverage for property damage.
Legal Guidance: Ensuring full compliance with local zoning laws and land regulations as we expand.

Technical & Business Roadmap: In Year 1, our goal is to acquire 250 hosts, process 5,000 bookings, and establish a B2B partnership. On the engineering side, we will migrate Stripe to asynchronous Webhooks for resilient payment states and introduce AWS SQS queues to offload background notifications. The problem exists, the solution exists, and we are ready to scale.