---
title: "Orhun Küpeli (@orhunkupeli) | Hashnode"
description: "Orhun Küpeli (@orhunkupeli) writes on Hashnode. Read their latest articles and follow their work."
url: "https://hashnode.com/@orhunkupeli"
---

@orhunkupeli

# Orhun Küpeli

@orhunkupeli · Munich, Germany · Joined May 2026

### [One GPU, Two LLMs](https://orhunkupeli.hashnode.dev/one-gpu-two-llms?utm_source=hashnode&utm_medium=feed)

Building a production-grade LLM inference platform on EKS with GPU time-slicing, KServe, Knative, and a custom gateway. Finally, debugging everything that broke.

6 min read · [orhunkupeli.hashnode.dev](https://orhunkupeli.hashnode.dev/one-gpu-two-llms?utm_source=hashnode&utm_medium=feed)

[Jul 21](https://hashnode.com/posts/one-gpu-two-llms/6a5fb8e7acded6fb89b82065)

### [The Numbers: Benchmarking My LLM Gateway on a H100](https://orhunkupeli.hashnode.dev/the-numbers-benchmarking-my-llm-gateway-on-a-h100?utm_source=hashnode&utm_medium=feed)

Real H100 benchmark of a production LLM gateway vs a naive HuggingFace baseline continuous batching, streaming TTFT, graceful 429s.

5 min read · [orhunkupeli.hashnode.dev](https://orhunkupeli.hashnode.dev/the-numbers-benchmarking-my-llm-gateway-on-a-h100?utm_source=hashnode&utm_medium=feed)

[Jun 21](https://hashnode.com/posts/the-numbers-benchmarking-my-llm-gateway-on-a-h100/6a37c1660984820a823b1126)

### [MVP to Mission-Critical: The Idea Behind My LLM Gateway Rewrite](https://orhunkupeli.hashnode.dev/mvp-to-mission-critical-the-idea-behind-my-llm-gateway-rewrite?utm_source=hashnode&utm_medium=feed)

How I rebuilt my LLM gateway from MVP to production: multi-tenant isolation, vLLM inference, budgets, and sub-5ms overhead. Architecture deep dive

5 min read · [orhunkupeli.hashnode.dev](https://orhunkupeli.hashnode.dev/mvp-to-mission-critical-the-idea-behind-my-llm-gateway-rewrite?utm_source=hashnode&utm_medium=feed)

[May 14](https://hashnode.com/posts/mvp-to-mission-critical-the-idea-behind-my-llm-gateway-rewrite/6a05c3b7baf09db7a6fcc626)
