CloakNotes · Private AI Workspace
On-device audio workspace running quantized INT4 LLMs with <1.1GB RAM active footprint and zero cloud latency.
I build low-latency, resilient systems on constrained hardware, from mobile to edge to embedded.
software
engineer
Featured
works
AI & Edge SystemsReal-Time Systems
Veloq · Real-Time Racing HUD
120 FPS sim-racing telemetry HUD with zero-allocation binary parsing and direct GPU canvas repainting.
ArokoDB · Embedded Vector Database
Embedded vector search engine delivering <12ms recall across 100k 768-dim vectors using Rust and mmap storage.
Stralis · Media Automation Engine
Distributed media automation engine compiling complex multi-step pipelines into resilient, fault-tolerant DAGs.
CowryPod · Digital Wallet
High-concurrency micro-transaction ledger processing 18,000+ TPS using append-only event sourcing and Redis batching.
Vezati · Field Operations Platform
Offline-first field operations client with vector clock conflict resolution and viewport-based CAD tile rendering.
// all worksexplore more
I turn hard constraints into measurable performance: on-device AI, mobile runtimes, edge infrastructure, and local-first architecture.
- 01
I build for the constraint
quantized models on phone hardware, runtimes that respect memory, systems that stay fast when the network disappears.
- 02
I break on purpose
safely and deliberately: the same instinct that finds the fault already knows the fix, so I adapt fast and never panic.
- 03
I write for the next engineer
code other people can actually pick up and work with: clear, documented, and built to hand over.
Tools for serious systems
Selected stackCore Systems & Runtimes
C++C++RSRustDartDartKKotlinkernelLinux KernelWASMWebAssembly
Edge AI & Local Inference
LiteRTLiteRT (TFLite)ONNXONNX RuntimeHNSWHNSW Vector SearchSIMDSIMD (NEON/AVX2)
Distributed Infrastructure
PGPostgreSQLRedisRedisDockerDockerK8sKubernetesTFTerraformgRPCgRPC / Protocol Buffers