Vyay is a high-velocity expenditure audit utility designed for engineering leadership and founders to rationalize AI infrastructure costs. Delivering actionable financial intelligence in under 60 seconds, Vyay identifies service redundancies and plan sub-optimality through a 100% deterministic analytical engine.
Live Platform: https://vyaytarunya.vercel.app/
During the development of Vyay, I prioritized Time-to-Value and Institutional Defensibility. Below are the 5 core trade-offs made:
- Deterministic Logic vs. LLM Reasoning: I intentionally avoided using AI for the audit math itself. Financial audits require 100% repeatability. By using a hardcoded rule engine based on verified vendor data, we ensure that every recommendation is mathematically sound and CFO-ready.
- Zero-Authentication Model: To maximize lead generation velocity, I bypassed traditional sign-up walls. The friction of account creation is the primary bounce factor for founders; Vyay provides insight first and requests identity (email) only after value is demonstrated.
- Client-Side Orchestration: The audit engine executes entirely on the client. This offloads computational overhead from our infrastructure, ensuring near-infinite scalability and sub-100ms execution times without increasing OpEx.
- Zustand Persistence: I chose Zustand with LocalStorage persistence over database-only state. This allows users to leave and return to their specific audit progress without needing an account, maintaining a high conversion rate in a multi-step wizard.
- Calm Engineering Aesthetic: I rejected flashy "AI gradients" and excessive animations in favor of a data-dense, industrial-minimalist design (Cream & Charcoal). This aesthetic establishes the "Institutional Trust" required for a financial tool.
- Deterministic Audit Engine: 100% rule-based reasoning for financial defensibility.
- Hybrid AI Synthesis: Gemini 2.5 Flash integration for human-centric executive summaries.
- Zero-Friction Persistence: Supabase-backed report storage with unique, shareable URLs.
- Lead Capture & Distribution: Transactional email integration via Resend for PDF report delivery.
- Operational Transparency: Comprehensive documentation covering GTM, Economics, and Architecture.
- Production Grade: 90+ Lighthouse accessibility scores and a robust Vitest suite.
| Document | Description |
|---|---|
| ARCHITECTURE.md | Mermaid system diagrams, data flow, stack justification, and scaling strategy. |
| DEVLOG.md | Day-by-day technical narrative and project progression log. |
| REFLECTION.md | Philosophical analysis on deterministic logic vs AI hype and technical reflections. |
| TESTS.md | Automated test suite documentation and audit engine coverage. |
| PRICING_DATA.md | Verified vendor pricing data with official citation URLs. |
| PROMPTS.md | Complete LLM prompts used, rationale, and iteration history. |
| GTM.md | Go-to-market strategy, user acquisition blueprint, and traction metrics. |
| ECONOMICS.md | Unit economics, LTV/CAC analysis, and $1M ARR roadmap. |
| USER_INTERVIEWS.md | Real-world insights from engineering leadership and founders. |
| LANDING_COPY.md | High-fidelity messaging, CTAs, and social proof orchestration. |
| METRICS.md | North Star metric, instrumentation plan, and pivot triggers. |
| PRODUCT_SPEC.md | Comprehensive product specification and implementation vision. |
| CI Workflow | GitHub Actions configuration for automated linting and testing. |
- Frontend: React 18, Vite, TypeScript, Tailwind CSS
- Intelligence: Gemini 2.5 Flash (Analytical synthesis)
- Data/Persistence: Supabase (PostgreSQL), Zustand (LocalStorage)
- Communications: Resend (Transactional Email)
- Infrastructure: Vercel (Edge Deployment)
# Clone the repository
git clone https://github.com/TarunyaProgrammer/vyay-aiCreditsAuditer.git
# Install dependencies
npm install
# Execute development environment
npm run dev
# Execute test suite
npm testDeveloped by Tarunya for the Credex Web Development Internship. "Calm technology for engineering budget recalibration."