Writing
All Articles
Design notes, engineering trade-offs, and lessons from building AI systems and data products.
Sub2API Gateway Internals and Enterprise Agent Architecture
A source-level study of streaming proxying, account-pool scheduling, atomic Redis concurrency control, multi-level authentication caching, and idempotent billing in Sub2API.
Knowledge Engineering and Tunable Cockpit Architecture for Enterprise RAG
A deep dive into enterprise RAG knowledge engineering: structure-aware chunking, hybrid Dense+Keyword retrieval with RRF reranking, lifecycle metadata governance, MCP boundaries, and the Retrieval Lab evaluation loop.
Branch Governance with Cherry-Pick in Multi-Factory Deployments
Managing multi-site code deployments: using single-path cherry-pick workflows to distill features into atomic trunk commits before promoting to factory deployment branches.
Enterprise AI Cockpit: RAG, Vector Search, and Real-Time Streaming
Engineering an enterprise AI cockpit featuring decoupled persistent storage, pgvector retrieval with lexical fallbacks, true end-to-end SSE streaming, and strict resource isolation.
Walk-Forward Validation and Out-of-Sample Testing for LLM Sentiment Factors
Practical feature engineering with LLMs in quantitative trading: structured event extraction, temporal alignment, exponential decay, and Walk-forward out-of-sample incremental testing.
Traceable Data Architecture for Cross-Border Trend Intelligence
Integrating multi-source data ingestion, deterministic category normalization, and a profit waterfall model to deliver an auditable daily decision-support system.