Writing
Articles
Engineering notes on requirements, architecture, delivery, and personal projects.
Layering an LLM Application
An application boundary for shared Q&A and knowledge capture across Telegram and web entry points.
A Human-Reviewed Knowledge Ingestion Pipeline
A staged workflow for turning conversations and other sources into reviewed knowledge-base updates.
RAG Architecture and Retrieval Improvements
The retrieval path, embedding and chunking choices, and a clear distinction between completed changes and proposed evaluations.
LLM Routing and Cost Control
Moving model choice behind an application-level router, with intent-based routing and decision logs.
Telegram Knowledge-Base Bot
An MVP Telegram bot with RAG answers, navigation commands, deployment automation, and session isolation.
From Engineering Changes to a Project Knowledge Base
A deliberately narrow structure for recording project changes, capabilities, architecture, and major decisions.
How My Use of AI Agents Is Evolving
From using agents for coding tasks to designing an engineering workflow that remains useful and maintainable.