Khushvi Bamrolia

Founder and product/architecture lead of CoffeeQL, a zero-dependency Rust query engine shipped as a WebAssembly-enabled npm package. Built a custom Lexer → Parser → Planner → Executor pipeline that federates PostgreSQL, MySQL, MongoDB, and Redis with in-memory hash joins. Supported Node.js, Deno, Bun, and browser environments; maintained 500+ weekly downloads and passing 500/500 automated tests. Also used the product as the data-access layer for AgileWriter V2 in production. Owned roadmap, open-source strategy, and go-to-market activities; selected for the Sarvam AI Startup Program.

Led technical direction for a 6–8 engineer cross-functional team from design through production release. Owned 0-to-1 architecture and productionization of a conversational AI system including dialogue-state management, context-window optimization, and symbolic + vector retrieval. Shipped an enterprise document-generation product (SSO, multi-panel UI, SharePoint navigation, hybrid TF-IDF + graph retrieval, and Azure OpenAI generation with tracked-change Word outputs via Azure AD SSO through Supabase). Designed reliability for the LLM layer using a multi-layer resilience gateway (fuzzy caching, context routing, request packing, and AIMD rate control) and documented the approach in ADRs. Built an agentic memory loop that learns from user corrections using an LLM-as-judge reflection pipeline, storing scoped rules in pgvector with RBAC/TTL decay state management. Architected an end-to-end V2 rebuild using SvelteKit + FastAPI gateway + Rust core services + Python AI services, with Qdrant/

Prepared 20+ students for technical interviews covering DSA, system design, and algorithms; supported interview readiness through practice and structured preparation.