LEARNING PATH · AI Engineering

Ship an AI Product

For indie hackers and founders shipping AI products solo.

Intermediate ~3h 30m7 lessons13 steps

The build → ship → grow arc for a one-person AI startup: the on-site lessons for the engineering, woven together with open-source tools — Inkwell, VolfPack, PromptPlay — for prototyping, finding your first users, and marketing a launch without a team.

  • Go from a prompt to a grounded, agentic AI product
  • Evaluate quality and keep inference cost in check
  • Find your first users and market a launch with open-source tools
  • Ship in public, solo, without a team
0 / 7 done · 0%
  1. HandbookNext up

    The Prompting Handbook

    Reliable model output is the foundation of any AI product.

  2. OSS Tool↗ OSS tool

    PromptPlay

    See how fast prompt → shipped artifact can be.

  3. AI System Design

    Design a RAG Pipeline

    Ground the product in your own data with retrieval.

  4. ToolTool · optional

    RAG Chunking Playground

    Tune chunking and watch retrieval quality move.

  5. AI System Design

    Design an AI Agent System

    Let the product take actions, not just answer.

  6. ToolTool · optional

    Tool-Schema Designer

    Design the tool interfaces your agent calls.

  7. Handbook

    The Agent Evaluations Handbook

    Measure quality before your users do.

  8. ToolTool · optional

    Context Budget & Cost Planner

    Keep token cost predictable as you scale.

  9. AI System Design

    Design an LLM Gateway

    Route across providers with fallback, limits and caching.

  10. AI System Design

    Design an LLM Inference Server

    Serve it reliably and affordably.

  11. OSS Tool↗ OSS tool

    Inkwell

    Find and reach your first users.

  12. OSS Tool↗ OSS tool

    VolfPack

    Market the launch without a video team.

  13. Handbook

    The Senior AI Engineer Interview Handbook

    Harden the engineering as you grow.

The build → ship → grow arc

Shipping an AI product solo means wearing every hat, so this path is deliberately end-to-end. It starts with the engineering — reliable prompting, retrieval to ground the product in your data, agents to let it act, and eval plus cost control so quality and spend stay in hand — then hands off to open-source tools for the parts founders forget: finding first users and marketing a launch without a team.

It's opinionated toward doing, not reading: prototype fast, ground the product, measure quality before your users do, and ship in public. The goal is a real one-person AI product, not a tutorial.

← All learning paths