RAG engineering: your problem is almost always retrieval quality
When a RAG system gives bad answers, teams blame the model. Usually the fault is upstream, in retrieval. Here is how to…
humaineeti contributor
Tag: Production AI
humaineeti contributor
One smart agent will not save you: why orchestration is the real moat
Models keep getting better and cheaper. What does not commoditise is deciding which agent acts, in what order, with which permissions. In…
humaineeti contributor
The new enterprise MCP spec moves the hard part onto your plate
MCP is going stateless and cloud-native with the 2026-07-28 release. That makes enterprise-scale agents possible, and it hands authentication, authorization and transport…
humaineeti contributor
AI agent evaluation: why the final answer is not enough
Grading an agent only on its final answer misses most of what can go wrong. Real AI agent evaluation looks at the…
humaineeti contributor
Evaluation as a service: should you build your eval stack or buy it?
Every team shipping LLM features eventually needs a real evaluation setup. The build versus buy question for evaluation as a service is…
humaineeti contributor
MongoDB Atlas hybrid search with $rankFusion: a practitioner’s guide
MongoDB Atlas now combines full-text and vector search in a single $rankFusion stage using reciprocal rank fusion. Here is how it works,…
humaineeti contributor
Buy, build, or orchestrate? How build is bypassing the 3C grid
The 3C model scores AI initiatives on Capability, Complexity and Criticality. The framework still holds, but one of its three axes has…
Chandra Sekhar Mukherjee