I have 15 years in marketing. Zero computer science background. Never opened a terminal before 2024.

And now I want to have the skills of a senior AI Engineer.

What brought me here

In the last few months I've built: a daily newsletter about China, an ideas marketplace, an analytics system, automation pipelines. All with "vibe coding" — I describe what I want and the AI writes the code.

Does it work? It works. But I'm a USER of AI. Not an ENGINEER.

The user calls the API and prays. The engineer understands WHY it works, WHEN it will fail, and HOW to optimize.

I'm done praying.

What a Senior AI Engineer DOES

1. Designs architectures (RAG, agents, pipelines)

2. Fine-tunes models for specific tasks

3. Deploys to production (latency, cost, scale)

4. Evaluates models with custom evals

5. Builds tool use, function calling, MCP

6. Optimizes prompts systematically

7. Understands embeddings, vector DBs, retrieval

8. Reads papers and implements them

9. Makes trade-offs with solid reasoning

10. Debugs ML problems

My advantages

  • ✅ REAL experience with LLM agents in production
  • ✅ Live product using AI (collection + editorial + fact-check pipeline)
  • ✅ Product and growth mindset
  • ✅ I live in China (DeepSeek, Qwen, Zhipu ecosystem)

What's missing

  • ❌ Fluent Python
  • ❌ ML fundamentals
  • ❌ Embeddings and vector search
  • ❌ Fine-tuning
  • ❌ Paper reading

The plan: 3 months

Month 1: Foundation

Python + ML basics via fast.ai + real RAG with China to Watch data + prompt engineering with method.

Month 2: Building

Agents from scratch + fine-tuning on the Mac mini + production infra + eval suites.

Month 3: Mastery

Multimodal + system design + Chinese ecosystem + capstone project.

Rule: every concept has a REAL project. No generic tutorials.

The final test

"Design the architecture of a system that collects news from 10 sources, deduplicates, ranks, generates a personalized summary, fact-checks, and sends — with observability, fallbacks, and cost < $50/month."

If I can answer with real trade-offs → mission accomplished.

The fear

I'm afraid of being forever a "marketer who tinkers with AI." Afraid that 15 years of marketing formatted me in a way that doesn't fit this world.

But marketing taught me to learn fast, test fast, and kill what doesn't work. That IS engineering. Just with a different name.

Let's see how it goes.

If you're in marketing and want to get into AI for real, follow along. I'll document everything — including the days I want to throw it all away.