| CARVIEW |
Roni Bhakta
Full-stack developer, passionate about building innovative Open Source solutions that solve real-world problems.
I build scalable, open-source systems empowering communities and solving real-world problems, impact-first.
Founder of Zonformat: Zero-overhead notation format cutting LLM token usage by approximately 50 percent versus JSON. Human-readable, lossless. Python and TypeScript libraries with schemas, streaming, validation. Over 99% retrieval accuracy; powers cost-effective GenAI. Repos gaining stars and downloads.
2025 GSoC Fellow turned contract dev at Internet Archive (Nov '25 to Jan '26): Own backend and APIs for Lenny, open-source lending for Library's. Enables affordable e-books, per Aaron Swartz's vision. Rolled OPDS feeds to over 100 beta libraries, CI/CD, observability. Co-created with Internet Archive's Open Library team; maintain pyopds2 library and Lenny monorepo (TurboRepo, TypeScript, React). Advancing secure, decentralized lending.
Stack: Python and FastAPI, MERN and Next.js, Docker and Kubernetes and AWS and Cloudflare, GenAI (LLMs, LangChain, embeddings, RAG) for end-to-end build, ship, ops.
Hiring full-stack for open infra or GenAI? DM @ronibhakta1 on X
Selected for Stanford University PhD research on AI coding tools & OSS collaboration
GSoC 2025 Fellow turned contract dev at Internet Archive
Founder of ZON Format — 50% LLM token reduction vs JSON
Impacted 11M+ users via Open Library
ZON: Zero-Overhead Notation Format
ZON (Zero-Overhead Notation) is a data serialization format designed to minimize token usage when working with LLMs. It reduces token consumption by approximately 50% compared to JSON while maintaining human readability and lossless data representation. Features include schema support, streaming, validation, and type coercion. Available as Python and TypeScript libraries with over 99% retrieval accuracy. Powers cost-effective GenAI applications and is gaining adoption with growing stars and downloads.
Lenny-app: Sandbox & Documentation for Lenny
Lenny-app is a sandbox prototype and documentation website for the Lenny lending system. It allows users and admins to interact with Lenny, try out the system with demo e-books powered by OPDS, and provides a landing page and documentation for developers. Built with TurboRepo (Monorepo), TypeScript, React.js, Next.js, Node.js, Express.js, and Docker, it supports both development and production modes. Lenny-app is designed to be future-proof, enabling easy integration, open-source contributions, and a great user experience.
PRS: Internet Archive’s Public Readium Service
PRS (Public Readium Service) is an internal project at Internet Archive to support Readium Manifest CLI and Thorium web reader integration. Contributed support for Readium Manifest CLI 0.4.0 and added Thorium web support. PRS enables secure, global lending of e-books via OPDS feeds and is designed to work with Lenny and other digital library systems.
Lenny: Open Source Lending System for Libraries (ArchiveLabs)
All around the world, "Little Free Libraries®" have been popping up on street corners and improving peoples’ lives, built on a simple principle: take a book, share a book. As our lives and books become digital, we’re losing the freedoms of ownership that make Little Free Libraries possible. Lenny is a new open-source, free, plug-and-play project that lets anyone, anywhere – libraries, archives, individuals – set up their own digital lending library online. \n Lenny is designed to be frictionless, affordable, and easy to adopt, enabling individuals and libraries to have their own public digital libraries.
As a GSoC 2025 Fellow at Internet Archive, I co-built Lenny from scratch, responsible for backend architecture, deployment, and API design. Collaborating with ArchiveLabs and the Internet Archive community, including Michael E. Karpeles (Mek).
Lenny is our answer to the vision of pioneers like Aaron Swartz: access to information is a right, not a privilege. Building Lenny strengthens the open source ecosystem and helps create a decentralized, global library for the future.
Learn more at lennyforlibraries.org and GitHub.
Nexxus: AI-Driven Microblogging Platform
Developed sentiment analysis for posts using LLM to categorize as neutral, positive, or negative and provide user recommendations for improved experience. Built the frontend with ReactJS, middleware with Hono (ExpressJS for Cloudflare), and PostgreSQL with Prisma accelerate for serverless pooling and Prisma ORM for data migrations. Integrated Azure for OpenAI resources and routes using Python. Currently enhancing user feeds to be more interest-specific using ML, LLM with Lang-Chain using Python for data processing and FastAPI for interaction, deployed on Azure. works inside a docker container with Nginx as a reverse proxy. The project is designed to be scalable and maintainable, with a focus on performance and user experience.
Code Club AGPIT Website
Leading the development of the official Code Club AGPIT website. Responsible for project architecture, feature planning, and team coordination. The website showcases club activities, events, and projects, and serves as a hub for student engagement and resources.
Uni Data Analysis
Created a web application for university staff to view detailed, year-wise student performance stats and generate dynamic reports with personal remarks on marksheets. Built an ExpressJS REST API using MySQL (XAMPP). Used Dynamic Programming for cumulative performance metrics. Included dynamic marksheet generation and an attendance system. Showcased an AI-powered ERP system Idea (AI-Driven Student Performance Analytics, which was a idea representation of a project (ERP) derived from this project at INDIA STARTUP MAHAKUMBH 24', receiving feedback from 10+ industry leaders.