Mustafaraza Mir — Senior iOS Developer & AI Engineer, Ahmedabad, India. 4.5+ years shipping Swift & SwiftUI products across healthcare, fintech, logistics and government. Now pairing native mastery with LLMs and on-device AI.
Client, company and personal products — designed, built and shipped end-to-end, from architecture to store review. Private work is described, never linked. Filter, then open any card for the full breakdown.
Personal builds, side products and company work summarised for public view. Private repositories are described, never linked — no source, no internal names.
Here is exactly how I add value to your product — today, not someday.
End-to-end Swift & SwiftUI apps built for scale. Shipped to thousands of real users across healthcare, government, logistics and fintech — architecture to App Store.
AI embedded directly in iOS — CoreML on-device inference, GPT and Claude API calls, Vision framework for real-time computer vision that makes apps genuinely smart.
Autonomous agents for task automation and intelligent workflows using OpenAI and Anthropic APIs — designed, built, deployed and monitored in production.
Code reviews, modular design, XCTest and CI/CD. I build systems that scale and stay maintainable — clean, documented, testable code every time.
BLE wearables, document scanning, real-time GPS, payment flows — battle-tested in production. Firebase, Stripe, Google Maps and Core Bluetooth fully integrated.
Full-site development from design to live deployment with v0, Cursor and Bolt. MVPs and client tools shipped at 10× the normal speed.
My latest open-source project — a token-optimisation framework for AI coding assistants, used from Claude Code to Cursor.
Universal token-optimisation framework and "specialist squad" for AI coding assistants. Cuts input tokens with AST blast-radius analysis, cuts output tokens by silencing model monologue, and ships a native MCP server plus role-based slash commands for review, planning, security, QA and shipping.
*Reduction figures as stated in the project README, measured on its own benchmarks. Star and download counts intentionally omitted — no vanity metrics.
SQLite symbol and call graph with BFS blast-radius calculation, so the model reads only the code a change actually touches.
Suppresses reasoning preambles and conversational filler. Instant code and diffs, nothing else.
Fan-in analysis, sensitive-file flags and a test-gap detector produce an objective 0–100 change-risk score.
/review, /plan, /security, /qa, /ship, /debug — each role runs on the exact blast radius of your diff.
Every company, every role — the actual problems I solved.
I research frontier models, build autonomous agents and deploy AI products end-to-end. That compounds every iOS skill I have.
Every iOS feature I build, I ask one question: how does AI make this ten times better? Deep native mobile expertise plus genuine AI fluency — that combination is rare.
Aug 2025 · System prompt design, chain-of-thought, Claude & GPT prompt structures
Aug 2025 · LLM interaction patterns, AI-assisted development workflows
Gujarat Technological University · 2019–2022