保留正文与补采附件。源页面当前未重新核验；具体限制见交付清单。

Revisit consent button We value your privacy We use cookies to enhance your browsing experience, serve personalised ads or content, and analyse our traffic. By clicking "Accept All", you consent to our use of cookies. Customise Reject All Accept All Customise Consent Preferences We use cookies to help you navigate efficiently and perform certain functions. You will find detailed information about all cookies under each consent category below. The cookies that are categorised as "Necessary" are stored on your browser as they are essential for enabling the basic functionalities of the site. ... Show more NecessaryAlways Active Necessary cookies are required to enable the basic features of this site, such as providing secure log-in or adjusting your consent preferences. These cookies do not store any personally identifiable data.

Functional Functional cookies help perform certain functionalities like sharing the content of the website on social media platforms, collecting feedback, and other third-party features.

Analytics Analytical cookies are used to understand how visitors interact with the website. These cookies help provide information on metrics such as the number of visitors, bounce rate, traffic source, etc.

Performance Performance cookies are used to understand and analyse the key performance indexes of the website which helps in delivering a better user experience for the visitors.

Advertisement Advertisement cookies are used to provide visitors with customised advertisements based on the pages you visited previously and to analyse the effectiveness of the ad campaigns.

Uncategorised Other uncategorised cookies are those that are being analysed and have not been classified into a category as yet. No cookies to display. Reject All Save My Preferences Accept All Toggle SearchToggle Mobile Navigation Publish AI, ML & data-science insights to a global community of data professionals. Sign inSubmit an Article LatestEditor's PicksNewsletterShipAI

Write For TDS LinkedInXToggle Search Data Engineering

From Data Analyst to Data Engineer: My 12-Month Self-Study Roadmap

The exact tools I'm learning, the projects I'm building, and the mistakes I'm already expecting to make Ibrahim Salami May 16, 20269 min read Generated with Gemini AI To be honest. A part of me started this journey because data engineering is one of the hottest and highest-paying careers right now. I'm not going to pretend that wasn't a factor. But there's more to it than that. I've been learning data analytics for a while now. SQL, Power BI, Python (Pandas, NumPy, a little Polars), data cleaning, EDA. You name it, I've been in the weeds with it. And I genuinely enjoy it. But somewhere along the way, I started getting curious about what happens before the data lands on my desk. How does it move? Who builds those pipelines? What does the infrastructure behind all of this actually look like? That curiosity planted a seed. Then AI started making a lot of what I do faster and easier. Which is great. But it also made me think: if AI can handle the analysis, what's my edge? What can I build and understand that goes deeper? I work as an IT System Analyst at a startup, and while I enjoy the work, I realized I wasn't challenging myself the way I wanted to. I was ready for more. The final push came from a video by Data With Baraa, where he laid out a complete data engineering roadmap. Something about seeing it structured and broken down made it feel real and doable. So here I am. I'm learning data engineering in public. And this article is the beginning of that journey. Also, just leaving a disclaimer that I'm not affiliated with Data with Baraa. I'm just sharing my personal journey. Hope it helps. Learn this step by step with the interactive Data Engineer roadmap.

Why Data Engineering Specifically

I want to spend a moment here because I think this question deserves a real answer. Data analytics taught me how to work with data after it arrives. Clean it, explore it, visualize it, draw insights from it. That skillset is genuinely valuable. But the more I learned, the more I kept bumping into the same wall. The data I was working with had already been shaped and moved by someone else. Someone had built the pipeline that brought it to me. Someone had decided how it was stored, how it was structured, how often it refreshed. I wanted to be that person. Data engineering sits upstream from analytics. It's about building the systems that make analysis possible in the first place. Data pipelines, storage architecture, workflow orchestration, large-scale data processing. These are the foundations everything else is built on. And honestly, that kind of infrastructure work appeals to me in a way that pure analysis no longer does. There's also a practical argument. Data engineering roles consistently rank among the highest paying in the data industry. As AI tools get better at automating the analytical layer, the demand for people who can build and maintain reliable data infrastructure is only going to grow. I'd rather be building the pipes than just using them. And one more thing. The startup I work at doesn't use any of the tools I'm about to learn. Which means every hour I put into this is entirely self-directed. No team to learn from, no work projects to apply it on. Just me, the internet, and whatever I can build on my own. That's a challenge I'm choosing on purpose.

Why I'm Doing This in Public

Writing about what I learn is something I already believe in deeply. It forces you to actually understand something before you explain it. It keeps you accountable. And over time, it builds something that a resume alone never could. But I'll be honest about my fears too, because I think that's the point of doing this publicly. I have shiny object syndrome. There, I said it. I've explored graphic design, animation, writing, marketing, and IT before landing in data. There's always something new and exciting pulling my attention. Data engineering could easily get replaced by the next flashy thing in my feed if I'm not intentional about it. Consistency is another one. I work a 9-5 where I barely touch the tools I'll be learning. There's no natural reinforcement at work, no colleague I can bounce Airflow questions off of. I'm building this entirely on my own time, outside of my job responsibilities. And balance. Three to four hours a day is the goal. Some days that will feel easy. Other days it will feel impossible. Publishing this journey is my accountability system. If I go quiet, you'll know I slipped. And I'd rather not slip.

What I'm Starting With

I'm not starting from zero, which helps. I already have beginner to intermediate SQL knowledge from my data analytics work, basic Python fundamentals, and some hands-on experience with Pandas. That gives me a foundation to build on rather than rebuild from scratch. Here's the full learning stack, roughly in the order I'll be tackling it.

1. SQL: Going Deeper Than Analytics

I know SQL. But analytics SQL and engineering SQL are different animals. I'll be going deeper into query optimization, indexing, working with very large datasets, and writing SQL that's built for performance rather than just exploration. If you've only ever used SQL to pull and filter data, there's a whole other layer underneath worth understanding.

Why it's first: Everything in data engineering eventually touches SQL. Getting sharp here before layering in more complex tools makes the rest of the journey easier.

2. Python: From Exploratory to Production-Ready

I have the basics. Pandas, NumPy, some Polars. But the Python I've been writing lives mostly in notebooks. Exploratory, messy, not built to last. The goal now is to write cleaner, more structured, reusable code. Functions, modules, error handling, scripting. The kind of Python you'd actually put in a pipeline.

Why it matters: Python is the glue that holds most modern data engineering stacks together. Airflow uses it. PySpark is built on it. Getting comfortable here is non-negotiable.

3. Git and GitHub: Version Control Done Properly

I'll be honest. My Git knowledge is currently "copy the command, hope it works." That has to change. Version control is fundamental to working like an engineer rather than just an analyst. I'll be learning branching, pull requests, and how to manage code properly across projects.

Why it matters: Every project I build from here on goes on GitHub. It's portfolio, it's discipline, and it's how real teams work.

4. Apache Spark and PySpark: Big Data Processing

This is where things get genuinely exciting. Apache Spark is one of the most widely used engines for processing large-scale data. PySpark is the Python API for it, which means I can use a language I'm already somewhat familiar with to work with distributed data at scale. The jump from Pandas to Spark is a mindset shift. Pandas works on a single machine. Spark is built to run across clusters. Learning to think in that distributed way is one of the skills that separates data engineers from analysts.

Why it matters: If you want to work with big data in a production environment, Spark is almost unavoidable. It shows up in job descriptions constantly and is core to the Databricks ecosystem I'll be building toward.

5. Apache Airflow: Orchestrating Data Pipelines

Data pipelines don't run themselves. You need something to schedule them, monitor them, and handle failures gracefully. That's where workflow orchestration tools come in, and Airflow is my pick. I considered a few options here. Databricks Workflows is great if you're already deep in the Databricks ecosystem. Azure Data Factory makes sense for Azure-heavy environments. But Airflow is free, open-source, cloud-agnostic, and widely used across the industry. It also teaches you the core concepts of orchestration in a way that transfers to other tools. Starting with Airflow felt like the right call, especially since I'm trying to keep costs low.

Why it matters: Orchestration is what turns a collection of scripts into an actual pipeline. Understanding Airflow is understanding how production data workflows are managed.

6. Databricks: The Data Platform

At some point you need to pick a data platform and go deep on it. I'm going with Databricks. It's built on top of Spark, it's in high demand, and it has a free Community Edition that lets you practice without paying for cloud credits. The alternatives are solid too. Snowflake is a clean, fast SQL warehouse that a lot of companies love. BigQuery is Google's fully managed, serverless option and genuinely excellent if you're leaning toward Google Cloud. But Databricks sits at the intersection of big data, machine learning, and data engineering in a way that matches where I want to go. It made the most sense for my goals.

Why it matters: Employers want you to have platform experience. Going deep on one is more valuable than knowing a little about all of them.

How I'm Structuring the 12 Months

The honest answer is that this might take longer than 12 months. And I'm okay with that. I'd rather take 15 months and actually understand what I'm doing than rush through in 12 and come out shaky on the fundamentals. The general approach is to move through each skill in order and not advance until I've built something with what I just learned. Tutorials are fine for orientation but projects are where real learning happens. My plan is to document each phase here on Towards Data Science: the concepts, the projects, the frustrations, and the wins. For tracking progress, I'm using the Notion roadmap from Data With Baraa as my backbone. It breaks down each skill into core topics and lets me track where I am without getting overwhelmed by the full picture all at once. As for time commitment, three to four hours a day is the target. Some of that will be structured learning. Some will be building. Some will be writing about what I just learned, which is its own form of studying.

What Success Looks Like

Landing a high-paying data engineering role is the goal. That's real and I'm not going to dress it up. But alongside that, I want to become a credible voice in this space. Someone who builds things worth talking about, documents the journey without filtering out the hard parts, and maybe makes the path a little clearer for someone coming up behind me. The writing and the learning feed each other. The portfolio becomes the proof. The proof builds the brand. That's the vision.

Starting Today

This article is my official start date. I'm not waiting until I feel ready or until everything is perfectly planned. I'm starting now, writing as I go, and letting the process be public and a little messy. If you're somewhere on a similar path. Whether you're in analytics thinking about engineering, in IT wondering what's next, or just someone trying to build skills that hold their value in an AI-accelerated world. Follow along. I think we'll have a lot to talk about. I'll also be sharing my learnings on my YouTube channel. So feel free to subscribe below and follow along. ··· This is the first article in an ongoing series documenting my data engineering journey. I'll be publishing regularly on my progress, the projects I'm building, and everything I learn along the way. And if you want to get access to the Notion template, in case you're on the same journey as I am, you can access it here. Follow along on my journey below. YouTube Medium LinkedIn Twitter

Register Now to Continue Reading

Unlock Towards Data Science with a Free Account

All articles on Towards Data Science are available to registered users. Create a free account today to access this article and thousands more. Register for freeSign in to your account

Written By

Ibrahim Salami

See all from Ibrahim Salami Career AdviceCareer ChangeData EngineerData ScienceProgramming Share This Article

Towards Data Science is a community publication. Submit your insights to reach our global audience and earn through the TDS Author Payment Program. Write for TDS

Related Articles

Deployment Models in Cloud Computing

Programming Understanding the difference between private, public and hybrid cloud Giorgos Myrianthous November 4, 20213 min read

Why Load Testing is Essential to Take Your ML App to Production

Machine Learning Load Testing a Flask ML Application using Locust Ram Vegiraju July 9, 20215 min read

Interactive Geographical maps with GeoPandas

Data Science Explore the geospatial data with interactive visualizations Parul Pandey November 10, 20217 min read

Apache Spark for Data Science - How to Install and Get Started with PySpark

Machine Learning Install PySpark locally and load your first dataset - Only 5 minutes required. Dario Radečić April 7, 20226 min read

Radical Simplicity in Data Engineering

Programming Learn from Software Engineers and Discover the Joy of 'Worse is Better' Thinking Cai Parry-Jones July 26, 20246 min read

From Cartography to Digital Mapping: Do you know the truth?

Data Science An online exhibition reflects on how the history of persuasive cartography can teach us to spot biases in visual data Nahal Sheikh September 27, 20207 min read

Latest picks: How to measure interpretability?

Machine Learning Your daily dose of data science TDS Editors November 23, 20201 min read

Working with Python and JSON

Data Science The Basics every Data Enthusiast have to know about As a data analyst, data engineer... Christianlauer May 25, 20214 min read

Building a User Event Data Pipeline

Data Science It's never been easier to take control of your data with a custom data pipeline Jeremy Dorn July 14, 20215 min read

Your home for data science and AI. The world's leading publication for data science, data analytics, data engineering, machine learning, and artificial intelligence professionals. Subscribe to Our NewsletterWrite For TDS•About•Advertise•Privacy Policy•Terms of Use © Insight Media Group, LLC 2026•The New Stack•roadmap.sh•Creator Network

Recommended Articles

Close