Vaibhav Dabhi

About

Solo builder of ZenDSA, a live AI-powered DSA learning product with 37 real users, built end to end using Java/Spring Boot, React, and TypeScript. Particularly interesting for teams building AI products: they designed a production LLM fallback architecture, enforced structured JSON outputs, monitored parse-failure regressions, and fixed an SSRF vulnerability after launch.

Key Strengths

• Built and deployed a full-stack AI product end to end
• Strong Java/Spring Boot and React/TypeScript implementation
• Designed resilient two-call LLM architecture with model fallback
• Structured-output prompting and validation for LLM reliability
• Production-minded security focus with SSRF vulnerability fix
• Pragmatic monitoring using parse-failure and fallback-rate signals
• Able to ship independently in ambiguous zero-process environments
• User-centered product thinking for learning workflows
• Live product traction with 37 real users
• Designed resilient LLM inference with circuit-breaker failover across multiple providers/models
• Reduced LLM token usage/cost by introducing difficulty-based token budgets (easy/medium/hard)
• Iterated prompts using explicit rules/constraints and user feedback to improve output quality
• Built an in-app feedback loop and used it to drive ongoing refinements
• Delivered a production feature with reported satisfaction from 38 active users

Experience

Education

Skills

Languages

Certifications