From Zero to Hired: A Student's Journey (Real Story)

When Maria K reached out to us in early 2024, she introduced herself as "a 29-year-old marketing manager who has no idea what she's doing." She had spent five years building brand campaigns for a mid-size consumer goods company in Munich. She was good at her job. She wasn't unhappy. But every time she read about AI reshaping industries, she felt the same nagging thought: I need to be on the building side of this, not the receiving side.

Seven months after that first email, she accepted a junior AI developer role at a Munich-based fintech startup. Starting salary: €52,000. No CS degree. No bootcamp diploma. Just a portfolio, a willingness to be a beginner for a while, and a plan she stuck to.

This is her story. She gave us permission to share it in full because, as she put it: "I wish I'd read something like this before I started. I almost talked myself out of it three times."

Month 1–2: Python Basics — The Uncomfortable Beginning

Maria started with zero programming experience. Her first two months were dedicated entirely to Python fundamentals: variables, loops, functions, lists, dictionaries. She spent about 90 minutes each evening after work, working through exercises on a structured course.

The first month was genuinely hard. "I kept hitting walls with object-oriented programming," she told us. "Classes, inheritance, self — none of it clicked intuitively. I spent three evenings just on understanding what 'self' meant in a method. Looking back, I was overthinking it. But at the time it felt like a sign that I wasn't cut out for this."

She almost quit in week three. What kept her going was one small win: a script she wrote to automatically rename a folder of photo files based on the date they were taken. "It was 15 lines of code and it did something real. Something I'd been doing manually for years. That was the moment I understood why people love programming."

By the end of month two, she was comfortable with Python basics and had built three small projects: the file renamer, a simple quiz game, and a script that fetched weather data from an API and printed a formatted morning report. Not impressive by developer standards. But real, working code she had written herself.

Month 3–4: Discovering Generative AI — The Field That Changed Everything

In month three, Maria stumbled on a YouTube video explaining how GPT works. She stayed up until 1am watching the follow-up videos. "I remember thinking — this is what I want to build with. Not ML models that take months to train. I want to build things that talk, that reason, that help people."

She pivoted her learning toward the OpenAI API and LangChain. Within two weeks, she had built her first chatbot — a simple FAQ bot that answered questions about Munich's public transport system. It wasn't perfect, but it worked. You could type a question and get a reasonable answer. She showed it to a friend who had no technical background. The friend was genuinely impressed. That mattered more than Maria expected.

Month four was about depth. She went beyond simple API calls and started understanding how LLMs actually work: tokenisation, context windows, system prompts, temperature settings. She learned about prompt engineering — not as a trick but as a craft. She started reading AI papers, not to understand the maths, but to understand the ideas.

"Month four was when I started feeling like I belonged in this field," she said. "Not because I knew everything — I definitely didn't — but because I was thinking about the right problems."

Month 5–6: The Portfolio Sprint — 40 Job Applications

By month five, Maria had a decision to make. She could keep learning indefinitely, or she could start applying and learn the rest on the job. She chose to apply — but first, she needed a portfolio that could speak for itself.

She spent six weeks building three projects with the deliberate goal of impressing a hiring manager:

Project 1 — DocChat: A RAG application that let users upload a PDF and ask questions about it. She used LangChain, ChromaDB, and Streamlit for the interface. Deployed on Hugging Face Spaces. This project alone got her three interview callbacks.

Project 2 — SentimentTracker: A tool that pulled recent tweets about a given company and ran sentiment analysis using a fine-tuned BERT model. Showed results in a Streamlit dashboard with charts. Demonstrated she could work with ML models, not just LLM APIs.

Project 3 — EmailDraft AI: A simple web app where you described an email you needed to send and got three draft options at different tones (professional, friendly, assertive). Built with React (which she learned specifically for this project) and a FastAPI backend. Took her three weeks to build. She was proud of it.

She then applied to 40 jobs in six weeks. Every application included her GitHub link and a one-paragraph description of her most relevant project. She tracked every application in a spreadsheet: company, role, application date, response, outcome.

Of those 40 applications: 12 companies never responded, 14 sent rejection emails, 8 invited her to a phone screen, 4 took her to a technical interview, 2 gave her a take-home project, and 1 made an offer.

Month 7: The Offer — A Junior AI Developer Role in Munich

The offer came from a fintech startup called Verdixx (name changed), a 60-person company building AI tools for financial advisors. The role: Junior AI Developer. Responsibilities: building and maintaining LLM-powered features, working with the backend team on API integrations, and owning the RAG pipeline that powered their document-analysis product.

The technical interview had three parts: a 30-minute conceptual conversation about how RAG works and how they'd improve it, a live coding exercise in Python (parsing a JSON response and handling edge cases), and a system design question about how to build a feature that lets users ask questions about their financial documents.

"The third question was basically what I'd built in DocChat," Maria said. "I didn't just answer abstractly — I told them exactly how I'd implemented it, what problems I'd run into, and how I'd solved them. They were nodding the whole time. I think that's when I knew I had the job."

She accepted at €52,000 with a six-month performance review and a clear path to €58,000 if she met her targets. She started two weeks later.

Three Lessons Maria Wants Every Career-Changer to Know

When we asked Maria what she'd tell someone starting the same journey today, she gave three answers without hesitating:

1. Don't wait until you feel ready. "I applied for the first 20 jobs before I felt confident enough to. I got rejected by all 20. But that feedback made me better. If I'd waited until I 'felt ready', I'd still be waiting. You only feel ready after you've done the thing, not before."

2. Your projects matter more than your CV. "Nobody at Verdixx cared that I'd been a marketing manager for five years. They cared that I had a working RAG app they could click around in and ask questions of. Build something real. Even something small. A live demo beats a hundred bullet points on a CV."

3. Community helps more than you expect. "I joined a Discord server for AI developers about halfway through my learning journey. I was terrified to ask questions because I thought I'd sound stupid. The first time I asked for help on a LangChain error I'd been stuck on for three days, someone answered in 20 minutes. Three days of frustration, solved in 20 minutes. The community is there. Use it."