Project Overview
Vizuna is an AI-powered workplace coaching platform designed to help employees reflect, receive behavioral insights, gather team feedback, and improve workplace performance.
The initial product was developed rapidly using AI-assisted coding tools to validate the concept. Our team was engaged to clean up and refactor the inherited AI-generated codebase, strengthen security, improve reliability, and transform the prototype into a scalable, enterprise-ready SaaS platform.


Challenges
The existing prototype had several technical and security limitations:
- Exposed API credentials and direct client-to-LLM API calls created security and compliance risks.
- Sensitive employee and enterprise data was being sent to AI services without proper redaction.
- Multi-tenant database access lacked strict Row Level Security, increasing the risk of cross-organization data exposure.
- Poor session and prompt management caused context loss during coaching interactions.
- Unhandled application errors resulted in frequent crashes and inconsistent user experiences.
- Limited automated testing made it difficult to identify regressions and safely introduce new features.
Our Solutions
We implemented a complete Vibe Coding Cleanup and modernization process across the platform.
- Removed client-side API credentials and introduced secure backend proxying using Supabase Edge Functions.
- Added custom PII redaction and token rate-limiting before sending data to external AI models.
- Implemented and validated strict Row Level Security policies across the PostgreSQL database.
- Refactored Android, iOS, and Next.js application layers to improve code quality and maintainability.
- Introduced structured context management to preserve reflection history and improve AI coaching continuity.
- Built automated regression tests for API security, database permissions, error handling, and core user workflows.
- Strengthened CI/CD workflows using automated testing and code-quality validation.

Our Results
The cleanup and modernization delivered measurable improvements across security, performance, reliability, and development efficiency.
- 100% client credential elimination, removing 3 exposed API keys and 12 direct client-to-LLM calls.
- 100% Row Level Security enforcement with 24 verified RLS policies across multi-tenant database tables.
- Fixed 14 security vulnerabilities, including high-severity PII exposure and authorization issues.
- Resolved 18 code-level defects across the application.
- Increased automated test coverage from 0% to 82%.
- Reduced application crash rate from 4.2% to below 0.1%.
- Improved average AI response time by 35% through context payload optimization.
- Eliminated PII exposure from outbound AI prompt payloads.
- Reduced developer onboarding time for new features by approximately 40%.
The result was a secure, stable, and maintainable enterprise SaaS platform ready for scalable adoption and future product development.
What Our Client Say’s
Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry’s standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged.
Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry’s standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged.





