Kazi Musharraf — AI Engineer & Developer Tools Portfolio
AI and developer-tools work — LLM apps, agent systems and automation utilities, built and deployed as real, usable products.
The emphasis is practical AI engineering: features people actually use, running in production, with the guardrails that make them dependable.
How I work
- Understand. I start with the real problem, users and constraints — not a feature list. A short discovery avoids building the wrong thing.
- Plan. A clear plan: scope, non-goals, the approach, and how we'll verify it works. You always know what's being built and why.
- Build. Small, tested, verifiable increments. Working software early and often, with clean, maintainable code that follows the project's conventions.
- Ship & verify. Real deployment, real testing, honest reporting. I verify behavior end-to-end before calling anything done — no 'it should work'.
Tech stack & tools
Models
- Claude
- OpenAI
- Groq
SDKs
- Vercel AI SDK
- AI Gateway
- MCP
Patterns
- Agents
- RAG
- Vision
- Structured output
Selected work
MK QuizFlow
An AI study tool that turns PDFs, notes and images into quizzes and flashcards. Next.js App Router, Vercel AI SDK (Claude), local-first IndexedDB storage, cloud share via Vercel Blob, an Image → Quiz vision feature, rate limiting and quota. Live at thequizflow.com.
MK Router
A time-based portfolio router that reverse-proxies five portfolios on one domain, rotating hourly. Shipped on both Cloudflare Workers and Vercel Edge Middleware.
60-project ecosystem
A large ecosystem of portfolios, tools, games and web apps, covering AI applications, developer tools and web projects.
MK GitFlow
An interactive Git simulator and learning platform — type real Git commands and watch an in-memory commit graph redraw live.
What you get when you hire me
- Clean, documented, maintainable code you own
- Live deployment with a working URL
- Tests for the important behavior
- Honest status reporting — what's done, tested, and any risks
- Clear handover so your team can keep building
Who I work with
Founders & startups
Building an MVP or first product and need someone who can ship the whole thing quickly, then iterate as you learn.
Small teams
You have a team but need extra capacity or a specialist to own a hard piece and deliver it end-to-end.
Agencies
You need a reliable contractor who communicates clearly, hits scope, and produces code your team can maintain.
Solo operators
You have an idea and a budget but no engineer — I can take it from concept to a live, working product.
Ways to work together
- Fixed-scope projects — a defined build with a clear deliverable and timeline
- Contract / part-time — ongoing capacity for a set number of hours
- Rescue & improvement — take over, stabilize and extend an existing codebase
- Advisory — architecture, code review and technical direction
I'm Kazi Musharraf (also known as Musharraf Kazi or MK Knight), a developer based in India working with clients worldwide, remote-first across time zones. Whether you need a one-off build or an ongoing engineering partner, the priority is the same: ship real, working software and report on it honestly.
Frequently asked questions
What's in this AI portfolio?
LLM apps, agent systems and developer tools that are live and usable.
Is it open source?
Much of the code is public on GitHub (mk-knight23).
What's the best example of your AI work?
QuizFlow — it turns PDFs and images into quizzes using LLMs and vision, with cloud sharing.
How can I hire you?
Use the contact or hire page with your project details.
Explore this portfolio
Explore project source links, engineering services or contact Kazi Musharraf.
How to assess my engineering work
My engineering case studies explain the user problem, architecture, source evidence and implementation limits. They include QuizFlow, MCP developer tooling and a typed LangGraph workflow. Review the linked code before evaluating fit for your project.
For AI features, I discuss input validation, structured outputs, credential handling and failure behavior. Production scale, accuracy and cost improvements require measurements specific to each system.