I build AI agents, automation pipelines, and the infrastructure that makes them safe to run. Software engineer by trade, Google Certified Cloud Architect, and deep in the agentic stack: Claude Code, MCP, n8n, and LLM-powered systems.

• 🛠️ Currently building warden, a security & observability gateway for AI tools
• 💬 Ask me about AI agents, automation, cloud architecture & system design
• 📫 Reach me at contact@parveshsaini.com · parveshsaini.com

• 🛡️ warden: Open-source MCP gateway that sits between AI clients and tool servers, adding prompt-injection & tool-poisoning detection, per-tool policy, rate limiting, approval gates, and OpenTelemetry tracing.
• 🖥️ AI-OS: A Claude Code starter kit that turns the CLI into a personal "AI Operating System". It onboards you through an interview, then runs recurring automation rituals.
• 🤖 personal-assistant: Aria, a voice- and text-driven AI assistant that connects to your calendar, email, Slack, and Notion and takes actions across all of them from one place.
• 🐦 twitter-backend: Production-style Twitter clone backend built on GraphQL (Apollo), Prisma/Postgres, Redis, and Kafka, with S3 media and full Prometheus + Grafana observability.
• 🔎 agentic_QA_system: An agentic RAG question-answering system on LangChain, augmenting a vector store with live web search (Tavily) to ground answers.
• 📢 claude-changelog: A bot that watches the official Claude Code changelog, summarizes each new release with Gemini, and auto-posts it to X.
• ⚙️ multithreaded-server: A multithreaded HTTP server written from scratch in Rust, backed by a hand-rolled thread pool. 🦀
• 🔀 dynamic-reverse-proxy: A TypeScript reverse proxy that talks to the Docker API to route requests and spin up backend containers on the fly.
• 💬 scalable-chat: A real-time chat app in a Turborepo monorepo, pairing a Next.js client with a decoupled Socket.IO gateway.
• 🎮 100boxes: A real-time collaborative 10×10 grid (r/place-style) where any cell you edit locks for 20 seconds for everyone.

• 🦞 openclaw/openclaw: Personal AI assistant that runs on any OS and any platform.
• 🔄 n8n-io/n8n: Fair-code workflow automation platform with native AI capabilities.
• 🦥 unslothai/unsloth: Faster, memory-efficient fine-tuning and training for open LLMs.
• 🔔 novuhq/novu: Open-source communication & notification infrastructure for agents and products.
• 🧠 topoteretes/cognee: Open-source AI memory platform that gives agents long-term memory.
• 🧰 langchain-ai/deepagents: A batteries-included harness for building deep, long-running agents.

Building in the open. Let's connect.