Android Architect · Android Artist · On-Device AI
I'm a Lead Android Engineer, Android Architect, and Android Artist (known online as @DearDhruv). I work on the parts of software that get interesting when they hit real device constraints: performance profiling, low-level hardware interfaces, and running local AI models on personal devices without cloud dependencies.
The technical areas where I spend most of my time, from native Android internals to cross-platform runtimes and low-level device hardware.
Production Clean Architecture, unidirectional data flow (MVI/MVVM), Jetpack Compose, modular multi-module codebases, memory profiling, and battery guardrails.
Cross-platform architecture sharing deterministic game loops, MVI state stores, and offline-first persistence across Android, iOS (SwiftUI host), and Android TV.
Local LLM inference and on-device diffusion studios (llama.cpp, LiteRT, Apple MLX) running on consumer phones with speculative drafting and defensive driver crash sandboxing.
Digital commerce flows, payment terminal protocols, and resilient hardware abstraction layers across 30+ POS peripherals with defensive IPC.
Real-world applications and native libraries engineered for performance, resilience, and privacy.
An offline-first multimodal AI workstation running local LLMs and diffusion models directly on mobile silicon with zero cloud telemetry. Features multi-engine runtimes (llama.cpp JNI, LiteRT, Apple MLX Metal), dual-model Speculative Turbo decoding (2x speedup), native fork() Vulkan crash isolation, on-device Bonsai 4B diffusion, and pure-KMP document parsing across 25+ formats.
A high-framerate arcade action game engineered in Kotlin Multiplatform and Compose Multiplatform for Android, Android TV, and iOS. Runs at 60-120 FPS using a shared deterministic physics loop, Boost and Overdrive speed surges, zero-allocation hot-path rendering, conflict-free Cloud Firestore schema v5 sync, and 29 progressive achievements.
A lightweight, privacy-first companion and app hub crafted specifically for Android TV and Google TV devices. Built with Kotlin and Jetpack Compose for TV, providing an instant 1-click gateway to launch the Google Play Store, an installed TV apps drawer with real-time remote search, storage health monitoring, 10 OLED dark themes, and 100% fluid D-pad navigation with zero ads.
A six-part architectural log documenting the realities of running local LLMs, image generation, and document processing directly on mobile hardware.
Architecting offline-first mobile inference, setting up speculative drafting with smaller models, and managing VRAM limits on mobile silicon.
How a simple 'attach file' button turned into an offline ingestion pipeline for 25+ formats using pure RFC 1951 Deflate and memory-mapped chunks.
Handling vendor Vulkan driver GPU segmentation faults by running inference inside a native fork() process sandbox with a 3000ms watchdog loop.
Synthesizing 4-step diffusion images on mobile GPUs using Google LiteRT and Bonsai Image Ternary 4B without thermal runaway.
Refactoring a monolithic mobile app into a multi-module MVI platform with clean offline boundaries, strict state stores, and zero cloud leaks.
Engineering dual-model speculative drafting for 2x decoding speed, isolating GPU driver segmentation faults in a native fork() sandbox, and streaming fault tolerance.
Find me as @DearDhruv across developer networks, or reach out directly. Always glad to talk through native Android architecture, Kotlin Multiplatform systems, on-device AI runtimes, or upcoming technical speaking opportunities.
