CS engineer & AI-native builder.
Agentic systems and full-stack SaaS,
shipped to production. Hyderabad, India.
Flagship · Open Source // 01
Norn is an open-source MCP memory server: two-stage conflict detection, a review UI, and a live fleet dashboard. Published to npm.
Explore the work
Proof over promises: shipped products, live users, and one world record.
Real products with real users, captured live from the running software. No mockups.
Persistent-memory MCP server for AI agents, published to npm. An open-source memory layer built on the Model Context Protocol: a two-stage conflict-detection pipeline (embedding-similarity gate plus NLI model), a conflict-review UI, and a live fleet-monitoring dashboard with SSE streaming and a real-time orbital agent map. 89/89 tests passing.
The AI placement coach for Indian engineers: a free adaptive skill diagnostic across DSA, aptitude, OS, DBMS, CN and verbal, AI mock interviews with instant scoring, a live voice coach for spoken practice, and company packs for TCS NQT, Infosys SP and Wipro NLTH.
A full café and restaurant SaaS: QR-code table ordering, Razorpay payments, realtime order updates, a staff order dashboard, and a multi-branch admin panel, wrapped in a dark-luxury design.
An agentic AI system for automated tender evaluation. It pipelines unstructured procurement documents through an LLM reasoning layer to surface structured insights, cutting manual review time by an order of magnitude.
Norn, from problem to published package.
Every session with an AI coding agent starts from zero. It forgets your stack, your decisions, and your conventions, so you re-explain the same project every day. Naive memory tools make it worse: they store contradictory facts and silently rot.
Built Norn, a persistent-memory MCP server in TypeScript. Every new memory passes a two-stage conflict gate: an embedding-similarity check, then an NLI model that separates contradictions from updates. Conflicts land in a review UI instead of overwriting silently, and a live dashboard streams the whole agent fleet over SSE onto a real-time orbital map.
Published to npm as @samad001z/norn-server and @samad001z/norn-core, with 89/89 tests passing. One command installs it into any MCP client, Claude Code and Cursor included, and the agent's memory survives across sessions.
Builds that sharpened the craft: mobile AI, computer vision, forensics, and client work.
A cloud-kitchen ordering platform built for a home bakery in Hyderabad. My first production deployment serving real end users: full ordering flow, menu, and checkout.
React Native AI health companion. Uses Gemini OCR to extract structured data from medical reports with Pydantic-validated extraction, a FastAPI backend, and doctor QR-share.
Cross-platform grocery price-comparison app. Expo/React Native front end, FastAPI backend, Supabase with pgvector for semantic search, Gemini API, and a fintech-aesthetic UI.
Computer vision model for early plant disease detection from a single leaf image. Engineered for smallholder farmers on low-bandwidth, low-latency field connections.
Multi-modal misinformation detection platform. Combines LLM reasoning, digital forensics, and rPPG-based deepfake detection to score the authenticity of media end to end.
Production corporate website for a construction firm. Mobile-first information architecture, a performance-tuned asset pipeline, and WhatsApp lead capture, shipped to a paying client.
Three lanes, one outcome: ideas that actually ship. Most projects touch all three.
LLM systems, agentic pipelines, MCP infrastructure, and voice AI that reason over real, messy data instead of demos.
Complete products: frontend, backend, payments, realtime, and deploy, built to hold up in the wild.
A separate muscle: high-CTR thumbnail and visual design for video and social.
I optimize for shipping. Everything I build runs under live constraints: real users, real payments, real failure modes. Production teaches what specs can't.
Most student work never leaves the demo. I publish mine: Norn, my persistent-memory server for AI agents, is open source and live on npm for anyone to install. That bias toward shipping in public is the through-line.
I'm pursuing my B.Tech in Computer Science Engineering at Neil Gogte Institute of Technology (2024 to 2027), after a Diploma in Computer Science at Pallavi Engineering College. My work runs from agentic LLM systems and MCP infrastructure to full-stack SaaS with payments and realtime.
Build, break, learn, ship, repeat.
The tools I reach for to take an idea from whiteboard to production.
A four-step loop from idea to something running in production.
Contributing engineer on the team that set a Guinness World Record at the world's largest agentic AI hackathon.
Best academic project presentation across the institute.
National-level hackathon participant; shipped under aggressive deadline constraints.
Cloud fundamentals and modern production deployment workflows.
Hands-on exposure to native mobile application fundamentals and architecture.
A side craft, kept separate from the engineering: six favorite high-CTR thumbnails. Hover to pause.
The usual questions, answered up front.
AI products, MCP tooling, full-stack SaaS, and production web apps. I like taking something from a rough idea to a deployed, live product that real people actually use.
Yes. I'm open to internships, freelance builds, and full-time roles. The fastest way to reach me is the contact form below or a direct email.
The whole product. Architecture, backend, frontend, payments, and deploy, with the AI layer woven in where it genuinely earns its place instead of being a bolt-on.
Python and TypeScript at the core: React and FastAPI on the app side, Supabase or Firebase for data, and LLM tooling plus MCP for the AI. I pick the leanest option that survives production.
Absolutely. I'm comfortable jumping into an existing repo, matching its conventions, and shipping under real constraints and deadlines.
Three ways to plug me into your build.
You have a product and want agents, RAG, or voice AI in it. I design the pipeline, build it, and ship it behind real auth and payments.
Proof: Cracked AI, TenderIQYou have an idea and no team. I take it from whiteboard to a deployed product: frontend, backend, payments, realtime, and an admin panel.
Proof: Van Lavino, Bytes of JoyYou need agents that remember, coordinate, or reach your tools. I build MCP servers and the glue around them; my own is live on npm.
Proof: Norn, 89/89 testsOpen to internships, freelance builds & full-time roles