darjan.dev
CloudsForgeCloudsForgeCase Study

I built an AI video workstation - design through distribution.

I recently built a desktop app that rents you a cloud GPU and configures everything in a couple of clicks - so you can run open-source AI video models at the provider's own rate.

Visit cloudsforge.comAvailable now · Windows · macOS · Linux · $49 one-time
CloudsForge - Studio · Job #3
CloudsForge - a finished text-to-video job with prompt, render parameters, and output
01The Problem

Running open-source AI is unnecessarily hard.

These models need data-center GPUs that cost $10k+ - hardware almost nobody owns. Renting one yourself means conda environments, CUDA mismatches, dependency conflicts, and multi-page install guides. The alternative is paying 2-3× markups to subscription tools. The capability is open and free - the path to it is the wall.

  • ERR_01

    no $10k+ GPU on hand

  • ERR_02

    conda environments

  • ERR_03

    CUDA mismatches

  • ERR_04

    dependency conflicts

  • ERR_05

    multi-page install guides

  • ERR_06

    2–3× GPU markups

02The Solution

Four steps. No terminal required.

CloudsForge connects to RunPod, shows real GPU prices, auto-configures the whole environment, and gives you a clean UI for the job queue, logs, and gallery. You pay the GPU provider directly. Zero markups, zero telemetry.

  • 01Connects to RunPod - real GPU prices, you pay the provider directly
  • 02Auto-configures dependencies, models, and environments
  • 03Clean UI for the job queue, real-time logs, and gallery
  • 04Local-first, zero telemetry, lifetime ownership
CloudsForge - Cloud GPUs
CloudsForge - choose a model, pick a GPU by real price, watch the instance set up live
03The Journey

Ideation to ship - solo.

01

Research & Design

Found the gap - no friction-free desktop client for open-source AI video. Mapped the competition (Runway, Kling, Pika, raw terminal), then designed the full desktop workflow in Figma around one idea: simplicity with control.

02

MVP Development

Built the desktop app in Electron + React. Wired up the RunPod integration, a job queue with real-time log streaming, and a model manager supporting Open-Sora v2, Wan2.1, and HunyuanVideo.

03

Launch & Ship

Set up cross-platform builds and code signing through GitHub Actions, shipped on Windows / macOS / Linux, and handled distribution and licensing via Gumroad. One-time $49 - no subscription, ever.

04Tech Stack

The full spectrum - one person.

Frontend, desktop runtime, cloud orchestration, build pipeline, payments. No handoffs, no gaps.

  • 01Electron
  • 02React
  • 03TypeScript
  • 04Node.js
  • 05RunPod API
  • 06GitHub Actions CI/CD
  • 07Gumroad
05Demo Reel

Motion-graphics demo - coming soon

A full walkthrough reel is in production. The frames behind this slot are real CloudsForge generations.

06What I Learned

What shipping a product taught me.

Technical

Desktop is its own discipline

Auto-update strategy, cross-platform testing, code signing, and dependency management are problems the web never makes you solve.

Technical

Cloud APIs are async-failure machines

Job queuing, provider rate limits, network drops, state sync between local-first and cloud - error handling becomes the architecture.

Product

Positioning is a feature

Being explicit about what CloudsForge is NOT - not a model, not a provider, not a subscription - did more for clarity than any UI polish.

Founder

Building is the smaller half

Going from freelancer to founder means support, roadmap calls, and positioning - the code is necessary but it was never the whole job.

Available now · Windows · macOS · Linux · $49 one-time

I built this end-to-end. Now I want to do it with your team.

Design, frontend, backend, DevOps, payments - I've carried a real product across all of it, solo. I'm available for founding roles, senior engineering, and select freelance work.