Writing
Notes from the work.
Long-form pieces on what I've learned shipping production systems — AI pipelines, Flutter architecture, and the harder-to-name things about working with AI day-to-day.
- 01
Scaling Sacred Texts: Building a High-Throughput Translation Pipeline with Gemma 4 31B and vLLM
A deep engineering walkthrough of the Yahuah AI pipeline — how I translated 60,000 verses with zero errors using a custom DAG, GPU mutex gating, and a three-layer noise filter that caught 7,300+ false positives.
vLLMGemmaDAGEngineering2025 - 02
Migrating an Offline-First Flutter App from State-Heavy to Data-Driven Architecture
The story of rewriting Horeb. Why I moved off Bloc-as-data-warehouse, why ObjectBox won over Hive/Isar/Drift, and how a cleaner architecture cut memory use by 64% on its own.
FlutterObjectBoxArchitecture2024 - 03
Building a Mental Firewall Against AI Dependency
I noticed I was outsourcing my thinking to ChatGPT. Seven habits I built to use AI as a tool without losing the muscle of reasoning for myself.
AIReflection2024 - 04
Flutter GenUI — A Closer Look
A research dive into Flutter's GenUI package — how AI-driven UI generation actually works under the hood, plus a custom YouTube CatalogItem and AI tool I built to extend it.
FlutterGenUIFirebase AI2024