Dhruv Patel
Android Architect | Kotlin Multiplatform | On-Device AI
Lead Android & Multiplatform Engineer with 14+ years crafting high-performance native architectures, distributed KMP systems, and privacy-first on-device AI runtimes.
Specialization & Technical Focus
Deep expertise across native mobile systems, cross-platform runtimes, and low-level edge compute.
Android & Native Architecture
14+ YearsProduction Clean Architecture, unidirectional data flow (MVI/MVVM), Jetpack Compose, modular multi-module codebases, memory profiling, and battery guardrails.
Kotlin Multiplatform (KMP/CMP)
ProductionCross-platform architecture sharing deterministic game loops, MVI state stores, and offline-first persistence across Android, iOS (SwiftUI host), and Android TV.
On-Device AI & Edge Compute
Offline-FirstPioneered local LLM inference engines (llama.cpp, LiteRT, Apple MLX) on mobile silicon with Speculative Turbo decoding, diffusion studios, and defensive driver crash isolation.
Commerce & Hardware Integration
EnterpriseArchitected high-throughput digital commerce flows, secure payment terminal interfaces, and resilient hardware abstraction layers across 27+ POS peripherals.
Selected Production Systems
Real-world applications and native libraries engineered for performance, resilience, and privacy.
AI Playground
An offline-first, on-device multimodal AI workstation for Android and iOS that turns mobile devices into self-contained inference engines. Features dual-model Speculative Turbo decoding (2x speedup), defensive fork() child process Vulkan crash sandbox with SIGKILL watchdog, and a zero-dependency KMP document engine supporting 25+ formats.
VelociHue
A fast-paced, high-framerate (60–120 FPS) arcade action game engineered in Kotlin Multiplatform shipping across Android, iOS, and Android TV. Features a pure-Kotlin deterministic physics loop, conflict-free Cloud Firestore schema v5 sync with 3-way union merge, 29-achievement progression engine, and instant Compose share card generation.
Beyond the Model: AI Playground Deep Dive
A comprehensive six-part architecture series documenting how to run local LLMs, image generation, and document processing directly on consumer phones.
Architecting offline-first mobile inference, Speculative Turbo decoding with draft models, and VRAM memory admission controls.
ArchitectureIsolating vendor Vulkan driver GPU crashes with a fork() child process sandbox, 3000ms watchdog loop, and SIGKILL fallback.
Synthesizing 4-step photorealistic diffusion images on mobile GPUs using Google LiteRT and Bonsai Image Ternary 4B.
Diffusion AILet's Build Something Exceptional
Open to discussions on native Android architecture, Kotlin Multiplatform systems, on-device AI integration, and technical speaking.