I take LLM-powered systems from prototype to production — agentic workflows, RAG pipelines, and multi-agent orchestration — and lead the teams that ship them.https://salmanzafar.dev/https://salmanzafar.dev/blog/hello-world/https://salmanzafar.dev/blog/hello-world/Who I am, what I build, and what I'll write about here: agentic systems, RAG, and shipping LLM software to production.Fri, 17 Jul 2026 00:00:00 GMThttps://salmanzafar.dev/blog/context-engineering-is-the-job/https://salmanzafar.dev/blog/context-engineering-is-the-job/Prompt wording is a rounding error next to what the model can see and when. The discipline that replaced prompt engineering, and how we practice it.Tue, 14 Apr 2026 00:00:00 GMThttps://salmanzafar.dev/blog/why-promptev-is-byok/https://salmanzafar.dev/blog/why-promptev-is-byok/Bring-your-own-key looked like a commercial handicap. It turned out to be the architecture decision that makes enterprise conversations short.Tue, 24 Mar 2026 00:00:00 GMThttps://salmanzafar.dev/blog/prompt-injection-is-an-agent-problem/https://salmanzafar.dev/blog/prompt-injection-is-an-agent-problem/A chatbot that gets injected says something embarrassing. An agent that gets injected does something. Defense notes from systems with tools.Tue, 17 Feb 2026 00:00:00 GMThttps://salmanzafar.dev/blog/rag-in-75-languages/https://salmanzafar.dev/blog/rag-in-75-languages/Building retrieval that works in Urdu, Arabic, and Tagalog as well as it does in English. Spoiler: the embeddings were the least of it.Tue, 20 Jan 2026 00:00:00 GMThttps://salmanzafar.dev/blog/building-multimodal-context-engine/https://salmanzafar.dev/blog/building-multimodal-context-engine/Building a multimodal context engine — scanned PDFs, screenshots, charts, audio — and why 'just extract the text' is where pipelines go to die.Tue, 16 Dec 2025 00:00:00 GMThttps://salmanzafar.dev/blog/mcp-changed-agent-integrations/https://salmanzafar.dev/blog/mcp-changed-agent-integrations/Before the Model Context Protocol, every agent-to-tool integration was a bespoke adapter. Notes from moving production systems onto MCP.Tue, 11 Nov 2025 00:00:00 GMThttps://salmanzafar.dev/blog/graphrag-when-its-worth-it/https://salmanzafar.dev/blog/graphrag-when-its-worth-it/We run both plain RAG and GraphRAG in production. A field guide to which questions actually need the graph — and which just need better chunking.Tue, 30 Sep 2025 00:00:00 GMThttps://salmanzafar.dev/blog/human-in-the-loop-is-not-a-checkbox/https://salmanzafar.dev/blog/human-in-the-loop-is-not-a-checkbox/Approval gates that people rubber-stamp are worse than no gates at all. Designing HITL that survives contact with busy humans.Tue, 19 Aug 2025 00:00:00 GMThttps://salmanzafar.dev/blog/cutting-llm-inference-cost/https://salmanzafar.dev/blog/cutting-llm-inference-cost/Model routing, caching, and context compression — in the order they paid off. No heroics, mostly plumbing.Tue, 15 Jul 2025 00:00:00 GMThttps://salmanzafar.dev/blog/evals-as-ci/https://salmanzafar.dev/blog/evals-as-ci/We gate every prompt and retrieval change behind an eval suite, the same way we gate code behind tests. Here's the setup, and what it caught.Tue, 03 Jun 2025 00:00:00 GMThttps://salmanzafar.dev/blog/shipping-rag-to-enterprise-users/https://salmanzafar.dev/blog/shipping-rag-to-enterprise-users/Chunking strategies and vector databases are the easy 20%. The hard 80% is access control, stale documents, and users who ask questions no document answers.Tue, 22 Apr 2025 00:00:00 GMThttps://salmanzafar.dev/blog/most-ai-agents-are-workflows/https://salmanzafar.dev/blog/most-ai-agents-are-workflows/The industry keeps calling everything an agent. Most production systems I've shipped are workflows with LLM steps — and that's usually the right call.Fri, 14 Mar 2025 00:00:00 GMT