Vaibhav Saini
TypeScript · Python · Next.js · FastAPI · PostgreSQL
Software developer working across web, backend, and AI. Each project has documented architecture decisions, real technical trade-offs, and honest engineering.
Vaibhav Saini
Delhi, India
Value snapshot
About
Focus areas
Building software across web, backend, and AI.
I build projects across TypeScript and Python — focusing on clean architecture, reliable APIs, and thoughtful data modeling.
Recent work includes an AI code generation platform (Buildra), a multi-tenant collaborative workspace (Orbit), and a SaaS cost optimization tool (StackAudit). Each project is built with authenticated workflows, structured data, and documented architecture decisions.
Projects
Web. Backend. AI. Systems built from the ground up.
Users describe web applications in natural language and receive working Next.js applications inside cloud sandboxes. Built with an AI agent loop using Gemini, event-driven pipelines via Inngest, and real-time token streaming via Server-Sent Events.
Buildra
AI-powered code generation platform
- Architected an AI agent loop using Gemini and Inngest that iteratively generates, validates, and corrects code across multiple files.
- Integrated E2B cloud sandboxes for secure code execution with real-time preview and file system exploration.
- Implemented Server-Sent Events for token-by-token streaming, reducing perceived latency by ~60% compared to batch responses.
- Built with Clerk authentication, Prisma ORM, and React 19 Server Components.
A multi-tenant task management platform with organization-level data isolation, tRPC for end-to-end type safety, and fractional indexing for O(1) drag-and-drop reordering. Designed with Drizzle ORM and PostgreSQL.
Orbit
Multi-tenant collaborative platform
- Designed a multi-tenant PostgreSQL schema with organization-scoped RBAC and denormalized orgId for query performance.
- Implemented tRPC procedures with middleware for end-to-end type safety, eliminating runtime type errors between client and server.
- Built fractional indexing for Kanban board task reordering — O(1) position updates without rewriting adjacent rows.
- Created an invitation system with secure org-scoped invite links and role assignment on join.
A SaaS tool that audits AI/software spending and generates 12-15 actionable optimization recommendations. Features a resilient LLM pipeline (Claude → OpenAI → template fallback), shareable public report URLs, and a lead-capture funnel.
StackAudit
AI-powered SaaS cost optimizer
- Built a rules-based audit engine that generates 12-15 specific cost optimization recommendations per report.
- Designed a resilient LLM summary pipeline with three-tier degradation: Claude → OpenAI → template fallback.
- Implemented shareable public report URLs with auto-generated OpenGraph metadata for social preview.
- Created a lead-capture funnel with email collection and report download.