Writing
Technical Deep Dives
Short previews from Blueprints & Bugs my LinkedIn newsletter on ML and AI systems, real build lessons, failures, fixes, and reusable patterns. Each preview links to the full article on LinkedIn.
Blueprints & BugsBuild logs on ML & AI systemsBiweekly on LinkedIn
Newsletter LaunchJanuary 31, 2026
Welcome to Blueprints & Bugs!
An introduction to Blueprints & Bugs: a build-log newsletter about taking ML and AI systems from prototypes to production, including the architecture decisions, failures, fixes, and reusable patterns behind real systems.
RAG SystemsPublished on LinkedIn
From PDFs to Persistent Knowledge
A practical look at moving beyond one-off PDF chat toward persistent knowledge systems, where documents become reusable, searchable context for better research, reasoning, and decision support.
LLM EvaluationFebruary 17, 2026
Evaluating LLM-based Systems Before Production
Shipping an LLM feature without evaluation is like deploying a payment system without tests. This deep dive breaks down retrieval quality, answer grounding, reliability metrics, thin eval sets, and release gates for production AI systems.
GraphRAGApril 2, 2026
Why Standard RAG is Broken (And How Graph Analytics Fixes It)
Standard vector-only RAG hits a semantic ceiling when facts, relationships, and directionality matter. This piece explains why graph-backed retrieval can reduce structural blindness and make AI answers more reliable.
Production DebuggingPublished on LinkedIn
When the Backend Had Data but the Dashboard Showed Zeros
A production debugging story about the gap between backend truth and frontend display, tracing why data existed in the system while the dashboard still showed zeros.
Want the next deep dive first?
If you want to talk through an AI engineering problem, a writing idea, or a delivery challenge, reach out and let's continue the conversation.
Contact Me