The tech stack behind Creatiz (and why we picked it for an AI MVP)

As part of building Creatiz (our AI tool for LinkedIn content creation), we had to decide on a stack that was fast to ship, flexible enough to adapt, and stable enough to handle multiple experiments with LLMs.

Here’s what we went with:

• Frontend → React (fast iteration, huge ecosystem).

• Backend → FastAPI (Python) (simple, async-friendly, perfect for plugging in AI workflows).

• LLM orchestration → LangChain + LLaMA 3 (structured prompts + open-source flexibility).

• Model providers → Groq + OpenRouter (speed + access to multiple cutting-edge models).

• Dev tooling → Cursor + Claude Code (AI-assisted dev to ship faster).

💡 Why this stack?

• React + FastAPI → gives us a quick feedback loop and lets us separate concerns cleanly.

• LangChain + LLaMA 3 → modular experimentation without locking into one model.

• Groq + OpenRouter → helps us optimize for both performance and cost, testing which runs better for content tasks.

• Cursor + Claude Code → made early development 2–3x faster when iterating features and fixing bugs.

🚧 The Tradeoffs

• Running on LLMs can be costly (Claude credits in particular burned fast).

• Multi-tool orchestration adds complexity (LangChain + OpenRouter integration took time).

• Still learning what’s “just hype” vs what delivers real speed/quality for users.

For other founders/builders:

• If you’re building with AI, what stack did you choose?

• Any underrated tools we should try in our next iteration?

We’re in MVP mode, so we’re very open to swapping things out if it makes the creator experience smoother.

👉 If you’re curious, Creatiz is live here: https://creatiz.ai?utm_source=IndieHackers