Armin Mehran

Principal Engineer bridging the gap between mobile architecture and modern AI.

I bridge the gap between Enterprise Mobile Architecture and Modern On-Device AI. With over 10 years of experience, I build systems that are not only intelligent but also secure, scalable, and offline-capable.

📱 Privacy-First Edge Intelligence

Cloud latency and privacy concerns are the bottlenecks of modern AI. I specialize in Edge Execution—running Large Language Models (LLMs) and Computer Vision models directly on the Android device.

Project Spotlight: “EdgeGen” (Offline Chat Architecture)

  • The Challenge: Building a generative chat experience with zero data egress and 100% offline availability.
  • The Solution: Engineered a local inference pipeline using Google Gemini Nano and MediaPipe.
  • Result: Eliminated cloud costs and latency while guaranteeing user privacy.
  • Tech: Kotlin, Gemini Nano, MediaPipe, Coroutines.

Project Spotlight: Assistive Vision Pipeline

  • The Challenge: Real-time obstacle detection for visually impaired users where battery life is critical.
  • The Solution: Optimized a TensorFlow Lite model with CameraX, focusing on quantization to reduce thermal impact.
  • Tech: TFLite (Int8 Quantization), CameraX, Android Neural Networks API (NNAPI).

🏛 Enterprise-Grade Android Engineering

Before pivoting to AI, I spent years modernizing critical infrastructure for Tier-1 Financial Institutions. I don’t just prototype; I build production-ready software that meets strict compliance standards (OSFI).

Legacy Modernization & Architecture

  • Migration Expert: Proven track record re-architecting legacy Java monoliths into modern Kotlin and Jetpack Compose modules.
  • Performance: Experience reducing code size by 55% and optimizing build times for large-scale banking apps.
  • Reliability: Enforcing Clean Architecture (MVVM/MVI) and strict CI/CD pipelines (GitHub Actions) to ensure zero critical bugs in production releases.

đź›  Technical Stack

On-Device AI & ML

  • Models: Google Gemini Nano, MediaPipe, TensorFlow Lite, ML Kit.
  • Optimization: Model Quantization, TFLite Interpreter, Edge Inference.
  • Vision: CameraX, Real-time Object Detection.

Modern Android Development

  • Languages: Kotlin, Java, C++ (NDK).
  • UI Toolkit: Jetpack Compose, Material Design 3.
  • Architecture: Clean Architecture, MVVM/MVI, Hilt/Dagger, Coroutines/Flow.
  • Compliance: OSFI Security Standards, OAuth2, Biometric Auth.

This portfolio is hosted on GitHub Pages and emphasizes a “Code-First” approach to Engineering.