AI engineer or ML engineer? Which one are you?
November 10, 2026
AI literacy is turning into a baseline expectation across analyst, engineering, and business roles. What employers say is harder to find is the combination around it: someone who can build with AI, check what it produced, and explain the result to the people relying on it.
This week, we compare AI engineering and machine learning engineering, look at what is changing in coding jobs and data infrastructure, and share new datasets and references to build with.
Here's what's inside:
- Top Read: AI engineer or machine learning engineer? How to tell which part of the system you want to own. Compare the roles
- Dataquest Unlocked: six days left to access free Premium learning, win over the leaderboard, and earn five birthday badges. Keep your momentum
- From the Community: Shaping a report for the people who'll read it, and whether a certification beats a portfolio. Join the discussion
- Data & Resources: Satellite and digital-economy data, two references for checking your SQL and pandas code. Explore the resources
- What We're Reading: Why Python still leads as AI writes more code, data infrastructure built for agents, and a forecasting gap ML helped close. Start reading
Top Read
AI Engineer or ML Engineer? Pick the work, not the title.
Which part of the system do you want to spend your days on?
- ML engineer: Builds and owns the model
- AI engineer: Builds the product on top of it
Job postings use the two titles loosely, so this guide compares the work itself. ML engineers ask how to train a model that predicts well and stays accurate in production. AI engineers ask how to turn a model someone else trained into a product people can rely on.
The clearest part is a worked example: one support team, 5,000 tickets a week. The ML engineer trains a routing classifier and wrestles with inconsistent labels and drift. The AI engineer builds a RAG assistant that drafts replies and wrestles with citation accuracy, cost, latency, and prompt injection.
Why it matters: the same team, the same tickets, and two very different sets of problems to own. The guide also covers skills, salary ranges, and where to start from an analyst, data science, or software background. Both paths are open during Dataquest Unlocked, so you can try a lesson of each before Sunday.
Dataquest Unlocked
In full swing. Six days to make it count.
Dataquest Unlocked, day 4 of 9
Premium courses, career paths, skill paths, practice problems, and guided projects stay open through Sunday, November 15. If you've started a lesson, this is the week to turn it into a finished course or project. Everything you complete stays saved after the event.
Pick one goal → Keep your streak → Finish by end of Sunday.
- Climb the leaderboard: Points come from lessons, projects, and courses you complete, time spent learning, and your streak.
- Add new badges to your profile: Five birthday badges are up for grabs, each earned from what you do between November 7 and 15: First Step, Streak Keeper, Project Builder, Course Conqueror, and Pathbreaker.
- Bring a friend: Dataquest Unlocked is free to join, with no credit card needed. Send a friend your referral link and learn side by side for the last five days.
Power BI, Excel, and Tableau courses are excluded.
From the Community
See how other learners are putting the fundamentals to work, and making the case for the work they've done.
- Adapting Report Structure to Your Audience: Pastor explains why a data report should change its structure and flow for different decision-makers, while keeping the focus on the business outcomes that matter to them.
- Certification Exam vs. Project Portfolio: Martin weighs what a Google Cloud ML certification shows employers, and why a portfolio of real-world ML projects and cloud-deployed models shows even more.
Data & Resources
This week: two big sources of real data, two references for checking your own SQL and pandas code, and a framework for documenting how you evaluate an AI system.
- NASA Earthdata Search: Find and download satellite observations by topic, date, and geography, then build a geospatial, climate, or anomaly-detection project with real data.
- OECD Digital Economy and Society Outlook 2026: Use fresh indicators on AI adoption, compute, skills, privacy, and digital divides to build evidence-backed charts or a short policy brief.
- PostgreSQL 18 EXPLAIN reference: Profile a slow query with EXPLAIN and EXPLAIN ANALYZE, compare estimated and actual rows, and document how an index or rewrite changes the plan.
- pandas 3.0 migration guide: Upgrade a copy of a data-cleaning project and test string dtypes, Copy-on-Write, chained assignment, and datetime behavior before you migrate real code.
- NIST AI Resource Center: Turn an AI project into a documented evaluation workflow using the AI Risk Management Framework, its playbook, profiles, crosswalks, and testing guidance.
What We're Reading
This week: who is still hiring people who code, what data systems built for AI agents look like, and a forecasting problem machine learning helped crack.
- The most popular programming languages in 2026: With AI writing more code than ever, you might expect programming jobs to shrink. Drawing on nine industry surveys and US job posting data, this article finds Python still leads, and developer job postings have recently grown faster than the broader job market.
- What's new across Google's Agentic Data Cloud: Cross-cloud lakehouse caching, catalog federation, and direct Iceberg access from AlloyDB let agents use governed transactional and analytical data without repeated copies or ETL. Data roles increasingly need open table formats, catalogs, access control, and cost and latency reasoning alongside SQL and Python. (performance examples are vendor-selected)
- Better weather forecasts, two to six weeks out: Past about two weeks, weather has long been treated as mostly noise. MIT researchers built a machine-learning bias-correction layer that cleans up existing probabilistic forecasts instead of replacing them, and for the first time beat climatological averages in that range.