---
title: "Discussion on \"I Processed 190,000 Podcast Episodes on a GPU Server in My Apartment\" | Hashnode"
description: "Discussion on \"I Processed 190,000 Podcast Episodes on a GPU Server in My Apartment\". I built a GPU server at home to process 253,000 hours of audio. The lesson wasn't cheaper transcription — it was cheaper experimentation."
url: "https://hashnode.com/posts/i-processed-190-000-podcast-episodes-on-a-gpu-server-in-my-apartment/6a3afdd437b0a041a95a9159"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [I Processed 190,000 Podcast Episodes on a GPU Server in My Apartment](https://bjrn.hashnode.dev/i-processed-190-000-podcast-episodes-on-a-gpu-server-in-my-apartment?utm_source=hashnode&utm_medium=feed)

I built a GPU server at home to process 253,000 hours of audio. The lesson wasn't cheaper transcription — it was cheaper experimentation.

Jun 23 · 13 min read · bjrn.hashnode.dev  
Read the article

-   [#ai](https://hashnode.com/tag/ai)
-   [#self-hosted](https://hashnode.com/tag/self-hosted)
-   [#selfhosting](https://hashnode.com/tag/selfhosting)
-   [#startup](https://hashnode.com/tag/startup)
-   [#saas](https://hashnode.com/tag/saas)
-   [#asr](https://hashnode.com/tag/asr)

## Written by

[Ilya](https://hashnode.com/@bjorginson)

@bjorginson · Frankfurt am Main, Germany

  
[View profile](https://hashnode.com/@bjorginson)

Building AI-powered search infrastructure for large-scale podcast indexing: @PodZeus

Architect and deploy distributed systems · Run and manage inference on NVIDIA A30s · Build containerized worker pipelines in Go · Handle high-throughput media processing with ffmpeg &amp; RabbitMQ · Optimize Linux environments for performance, reliability, and cost.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
