InLevel Up CodingbyArti Dighe·5d agoSystem Design : An AI Tutor That Refuses to Do the WorkPart 1 of 7. The problem, the solution, the design, and the cost and latency calculations for an AI Tutor use case
InTowards AIbyArti Dighe·Nov 87 AI Use Cases Worth Trying in 2027Most of us still use AI like a search box. Here’s what to hand it instead in 2027, and how to start each one this week.A response icon2A response icon2
InTowards Data EngineeringbyArti Dighe·Oct 29Is It the Model or the Data? What I Learned When My AI Agent Gave Confident Wrong AnswersA beginner-friendly, step-by-step story about the one thing I now check first: what the AI actually sees.A response icon2A response icon2
InAI That ShipsbyArti Dighe·Oct 29Do I Need a Knowledge Graph? What I Learned Building an AI Agent for Amazon SellersA beginner-friendly, step-by-step story: from seller pain points, to a knowledge base, to the knowledge graph question.A response icon2A response icon2
InAI That ShipsbyArti Dighe·Oct 22Is MCP Dead? Where It Still Beats CLIs and SkillsWeigh the token costs, security lessons and latest spec changes before you decide which MCP servers to keep.A response icon1A response icon1
InData Science CollectivebyArti Dighe·Oct 17From Similarity to Connections: Enterprise Knowledge Graph RAG on Neo4j and AWSWhy similarity search fails on connected questions, and how we designed a graph-first RAG platform, including a safe migration off Amazon…
InAI That ShipsbyArti Dighe·Oct 8Top 4 Security anti-patterns for AI agentsFour designs that still ship. They are not exotic bugs. They are the defaults.A response icon1A response icon1
InData Science CollectivebyArti Dighe·Oct 8Top 4 Security design patterns for AI agentsFour patterns, enforced in code the model cannot rewrite. The model proposes. It does not decide.A response icon2A response icon2
InAI That ShipsbyArti Dighe·Sep 24Designing Evals for AI Agents: Why “It Passed” Isn’t Always “It Works”A green eval feels like proof. Five gentle questions show what a pass really promises, and how to make it tell the truth.A response icon1A response icon1
InData Science CollectivebyArti Dighe·Sep 11Designing the Agent Harness: Technical Approaches That Work in ProductionAn agent never fails the way a unit test fails. The loop keeps running, the tools keep firing — and the only sign of trouble is a task that…A response icon4A response icon4