Writing — Meleik Hyman https://www.meleikhyman.com/blog/ Long-form engineering notes by Meleik Hyman, Sr. AI/ML Engineer, on Graph-RAG retrieval, LLM evaluation, prompt chaining, chunking strategy, and hallucination mitigation. en-us melikehyman3@gmail.com (Meleik Hyman) melikehyman3@gmail.com (Meleik Hyman) Every stage of a RAG pipeline needs its own evaluation gate https://www.meleikhyman.com/blog/rag-evaluation-hooks/ https://www.meleikhyman.com/blog/rag-evaluation-hooks/ A single end-to-end score says the system got worse but never says which stage broke; per-stage gates turn a vague regression into a located one. Sun, 10 Aug 2025 00:00:00 GMT Meleik Hyman RAG Evaluation MLOps Graph-RAG needs eight evaluation gates, not one https://www.meleikhyman.com/blog/graph-rag-evaluation-hooks/ https://www.meleikhyman.com/blog/graph-rag-evaluation-hooks/ Graph-RAG fails in four places a vector index cannot, so extraction, graph construction, community detection, and summarisation each need their own measured gate. Sun, 10 Aug 2025 00:00:00 GMT Meleik Hyman Graph-RAG Evaluation Knowledge Graphs Prompt chains: deterministic pipelines that beat premature agents https://www.meleikhyman.com/blog/prompt-chaining/ https://www.meleikhyman.com/blog/prompt-chaining/ Fixing control flow at author time is what separates a prompt chain from an agent, and it buys predictable cost, reproducible traces, and testable steps. Sun, 10 Aug 2025 00:00:00 GMT Meleik Hyman LLM Prompt Engineering Architecture Sizing chunks so Graph-RAG extraction never loses an edge https://www.meleikhyman.com/blog/chunking-for-graph-rag/ https://www.meleikhyman.com/blog/chunking-for-graph-rag/ A four-tier routing rule that picks chunk size from document length, so every unit handed to the extractor can still yield at least one complete triple. Mon, 11 Aug 2025 00:00:00 GMT Meleik Hyman Graph-RAG Chunking Retrieval Five layers between a retrieval system and a fabricated answer https://www.meleikhyman.com/blog/mitigating-hallucinations/ https://www.meleikhyman.com/blog/mitigating-hallucinations/ Hallucination is several distinct failures under one name; five composable controls — grounding, instruction, verification, evaluation, guardrails — address them. Sun, 24 Aug 2025 00:00:00 GMT Meleik Hyman LLM Guardrails Evaluation Crossing the gap from Python script to double-clickable app https://www.meleikhyman.com/blog/shipping-stock-crypto-analyzer/ https://www.meleikhyman.com/blog/shipping-stock-crypto-analyzer/ A v1.0.0 release maps the three rungs between a working script and something a non-developer can open, and what each rung costs the person who built it. Fri, 03 Oct 2025 00:00:00 GMT Meleik Hyman Python Shipping Developer Tooling Pre-commit hooks are the feedback loop AI coding was missing https://www.meleikhyman.com/blog/pre-commit-ai-feedback-loop/ https://www.meleikhyman.com/blog/pre-commit-ai-feedback-loop/ An assistant that cannot see its own lint errors depends on a human to relay them; deterministic hooks turn that relay into a signal the machine can read. Fri, 09 Jan 2026 00:00:00 GMT Meleik Hyman Developer Tooling AI-Assisted Coding Code Quality