<![CDATA[Stories by Mayank Patel on Medium]]> https://medium.com/@maxy_ermayank?source=rss-5bc2e1c72c17------2 https://cdn-images-1.medium.com/fit/c/150/150/0*clk3SCjH5nmxXOz_.jpeg Stories by Mayank Patel on Medium https://medium.com/@maxy_ermayank?source=rss-5bc2e1c72c17------2 Medium Sun, 15 Nov 2026 03:00:16 GMT <![CDATA[AI Security & the LLM Threat Model: Why You Can’t Patch Prompt Injection]]>

An LLM reads untrusted text and can’t reliably tell instructions from data. Here’s how attacks get in — and how to keep the blast radius…

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/ai-security-the-llm-threat-model-why-you-cant-patch-prompt-injection-630641927b51?source=rss-5bc2e1c72c17------2 https://medium.com/p/630641927b51 Sun, 26 Jul 2026 18:38:03 GMT 2026-07-26T18:38:03.614Z <![CDATA[Evaluations, Done Right: How to Prove Your AI Is Actually Any Good]]>

Anyone can ship an AI feature that demos well. Evaluations are how you prove it’s good — repeatably, before your users find out it isn’t.

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/evaluations-done-right-how-to-prove-your-ai-is-actually-any-good-170ddd7781c1?source=rss-5bc2e1c72c17------2 https://medium.com/p/170ddd7781c1 Sun, 26 Jul 2026 02:07:09 GMT 2026-07-26T02:07:09.454Z <![CDATA[AI FinOps: Why Cheaper Tokens Are Making Your AI Bill Bigger]]>

Per-token prices are collapsing while AI bills keep climbing. The cost of an AI feature is decided in your architecture and code — not in…

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/ai-finops-why-cheaper-tokens-are-making-your-ai-bill-bigger-3a244d93a487?source=rss-5bc2e1c72c17------2 https://medium.com/p/3a244d93a487 Fri, 24 Jul 2026 02:54:39 GMT 2026-07-24T02:54:39.813Z <![CDATA[The LLM Gateway: One Control Plane for Every Model Call]]>

Routing, caching, fallback, budgets, guardrails, and observability — enforced once, instead of reinvented in every service.

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/the-llm-gateway-one-control-plane-for-every-model-call-d049a98971aa?source=rss-5bc2e1c72c17------2 https://medium.com/p/d049a98971aa Mon, 20 Jul 2026 22:59:00 GMT 2026-07-20T22:59:00.073Z <![CDATA[AI Observability: How to See Inside Systems That Never Behave the Same Way Twice]]>

Traditional monitoring tells you the service is up. AI observability tells you whether the answer was right — and lets you prove why.

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/ai-observability-how-to-see-inside-systems-that-never-behave-the-same-way-twice-da7084d76324?source=rss-5bc2e1c72c17------2 https://medium.com/p/da7084d76324 Mon, 20 Jul 2026 22:39:31 GMT 2026-07-20T22:39:31.417Z <![CDATA[Agentic Architecture: When the AI Acts, Not Just Answers]]>

A field guide to systems that pursue goals — the parts, the types, the orchestration, and what it takes to make them reliable.

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/agentic-architecture-when-the-ai-acts-not-just-answers-e8c1803af9ad?source=rss-5bc2e1c72c17------2 https://medium.com/p/e8c1803af9ad Thu, 16 Jul 2026 22:11:12 GMT 2026-07-16T22:11:12.756Z <![CDATA[RAG & Context Engineering: How the AI Actually Knows Things]]>

A model’s knowledge is frozen, lossy, and public. Retrieval and context engineering are how you feed it the right facts — yours, and…

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/rag-context-engineering-how-the-ai-actually-knows-things-ea8b4623f759?source=rss-5bc2e1c72c17------2 https://medium.com/p/ea8b4623f759 Wed, 15 Jul 2026 14:18:25 GMT 2026-07-15T14:18:25.213Z <![CDATA[How LLMs Actually Work: Tokens, Embeddings, and Why They Hallucinate]]>

No magic, no mystery — just a prediction engine. Here’s the pipeline from your words to its answer, and why a machine this fluent still…

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/how-llms-actually-work-tokens-embeddings-and-why-they-hallucinate-40d184243459?source=rss-5bc2e1c72c17------2 https://medium.com/p/40d184243459 Wed, 15 Jul 2026 03:55:56 GMT 2026-07-15T03:55:56.366Z <![CDATA[The LLM Iron Triangle: Why You Can’t Have Cheap, Fast, and Smart at Once]]>

A CAP-style trade-off every production AI system runs into — and how to bend it.

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/the-llm-iron-triangle-why-you-cant-have-cheap-fast-and-smart-at-once-805f68e46321?source=rss-5bc2e1c72c17------2 https://medium.com/p/805f68e46321 Wed, 08 Jul 2026 18:54:03 GMT 2026-07-08T18:54:03.530Z <![CDATA[Why Your Research Process Is Costing You $500 Per Report (And How AI Changes Everything)]]>

You’re a strategy analyst. It’s Monday morning. Your VP just Slacked you: “Need competitive landscape on EV battery suppliers by Thursday.”

Continue reading on Medium »

]]> https://medium.com/@maxy_ermayank/why-your-research-process-is-costing-you-500-per-report-and-how-ai-changes-everything-af7dfc7314a6?source=rss-5bc2e1c72c17------2 https://medium.com/p/af7dfc7314a6 Wed, 04 Mar 2026 05:14:22 GMT 2026-03-04T05:14:22.281Z