# Prayag Tushar

> AI Engineer at Metquay Inc. Based in India. i'm an ai engineer who builds llm-powered products — rag pipelines, multi-provider llm tooling, and the python apis behind them. i like turning messy problems into small, fast tools people actually want to use.

## About

- Name: Prayag Tushar
- Role: AI Engineer
- Company: Metquay Inc
- Location: India
- Site: https://prayagtushar.xyz
- Email: t.prayag.eng@gmail.com
- GitHub: https://github.com/prayagtushar
- X / Twitter: https://x.com/prayagcode
- LinkedIn: https://linkedin.com/in/prayagtushar
- Skills: Large Language Models, Retrieval Augmented Generation, LangChain, Vector Databases, Pinecone, pgvector, Embeddings, Prompt Engineering, LLM Evaluation, OpenAI, Anthropic, Gemini, Python, FastAPI, TypeScript, React, Next.js, NestJS, Node.js, AWS, PostgreSQL, Full Stack Development

## Pages

- [Home](https://prayagtushar.xyz): Bio, recent work, recent writing.
- [Blog](https://prayagtushar.xyz/blog): Long-form writing.
- [Work](https://prayagtushar.xyz/work): Employment history and what was shipped.
- [Projects](https://prayagtushar.xyz/projects): Side projects and open-source work.
- [Resume](https://prayagtushar.xyz/resume): Downloadable PDF resume.
- [RSS](https://prayagtushar.xyz/blog/rss.xml): Blog feed.

## Blog posts

- [picking a vector index without overthinking it](https://prayagtushar.xyz/blog/choose-the-right-vector-index): ivfflat vs hnsw vs streamingdiskann, in plain words.
- [Readora: Architecting a Modern PDF RAG Application](https://prayagtushar.xyz/blog/build-your-own-pdf-rag-app): A deep dive into building a production-ready Retrieval-Augmented Generation system using Next.js, Pinecone, and Gemini.

## Projects

- [Multi-LLM Client — Unified Async Client for OpenAI, Anthropic & Gemini](https://github.com/prayagtushar/multi-llm-client): A provider-agnostic async Python client that normalizes messages, streaming, token usage, and errors across OpenAI, Anthropic, and Gemini behind one interface. Ships as a library, CLI, interactive REPL, and FastAPI service — with tenacity retries, Pydantic v2 models, mypy-strict typing, a concurrent compare() across providers, and 40 tests.
- [Readora — RAG Chat Interface for PDFs](https://readora.prayagtushar.xyz/): Built a citation-style RAG chat app where users upload PDFs and ask grounded questions. Pipelines gemini-embedding-001 into Pinecone (namespace per file), streams answers via gemini-2.5-flash, and stores files on Vercel Blob with Neon Postgres + Drizzle ORM.
- [Ask Video.AI — SaaS RAG Bot for YouTube](https://youtu.be/-v93yz1Ik98): Developed a web tool for video querying using LangChain, Gemini API, and Pinecone. Handled transcript chunking and embedding storage, resolving data overlap issues for real-time responses. Managed codebase with Turborepo and Bun.
- [LumosAI — Chatbot Powered by Gemini](https://lumos-maxima.vercel.app/): Created a chatbot with Gemini API for conversations, tested over 100 sessions to refine response accuracy. Added Clerk login and MongoDB for chat storage. Used TailwindCSS for a mobile-friendly UI.

## Experience

- Software Engineer — Metquay Inc (June 2025 - Present)
- Software Developer — Zerone Consulting (Jan 2025 - May 2025)
- Application Developer — Collab Junction (June 2024 - Aug 2024)