Delivering a Disruptive Digital Addressing Platform for a High-Growth Startup

Solution

For this project, we assembled a powerful team including Front and Back-end Developers, Business Analyst, Quality Assurance Engineer and Project Manager with experience in building successful software solutions for different industries. 

We started with a comprehensive discovery phase to align the platform’s vision with market demand and user needs. An extensive audit of the existing VayaPin system revealed critical technical debt. The codebase showed signs of aging, including outdated libraries and inefficient architecture patterns that limited scalability. Based on discovery insights and audit findings, a phased product roadmap was developed to prioritize core features, technical improvements, and user experience enhancements. 

The development phase began with refactoring a significant portion of the legacy codebase, replacing over-engineered design patterns with a modular, service-oriented architecture that is easier to maintain and extend. 

To improve the monetization of a product, we implemented a tiered subscription system with role-based access and premium feature unlocks. Stripe is integrated as a primary payment system, fully connected to the user management system and billing logic on the backend.

For B2B enablement, we built and documented a suite of RESTful APIs and are currently developing an SDK to allow third-party platforms to integrate VayaPin capabilities directly into their services.

We also introduced multi-Pin management, allowing users to create and manage multiple VayaPins under a single account. Additionally, we designed a dedicated business dashboard, providing businesses with a centralized hub to manage their location-based digital presence.

A directory is only worth what’s in it, so alongside the platform rebuild, SPD built a second layer for VayaPin: an automated business-discovery and enrichment pipeline, plus a cold-email outreach system built on its output, inviting each business to claim and verify its auto-generated listing. Both were built with Claude Code, running primarily on Claude Sonnet 5, which writes and reviews the code at the engineering layer rather than appearing as a customer-facing feature.

A naive per-city search hits Google Places’ result cap silently, so the pipeline instead resolves a location to a real area, lays a recursive grid over it, and splits cells when they saturate the pagination window or cross the area’s boundary. Saturated cells too small to split get pushed through five more axes (type, price, rating, sort order, name tokens), and coverage is measured via an independent second pass rather than assumed.

Each business is enriched before publishing – website/phone via Place Details, coordinates via geocoding, email via a search-and-extraction chain – with a single model call generating subtitle, description, SEO metadata, and marketing copy in seven languages (coordinates never model-sourced, no unverifiable claims). On outreach, a send ledger keyed on Place ID plus campaign prevents duplicate emails across reruns and prior history, consolidated from a previous account. The system runs four campaigns, each a seven-email sequence over 30 days.