Teaching in the Age of AI

This fall semester, I taught Embedded Systems as a technical elective for fourth-year computer engineering students. 

Since most homework and many projects are now trivially solvable by AI (Large Language Models, LLMs), they are no longer worth the instructor’s effort to assign and grade. Therefore, I did not assign any homework. Instead, I encouraged students to use LLMs during class as a tool for active learning.

For each topic, I presented a short, real-world embedded example and then posed a question, such as “calculate the amount of stack used by the given C function.” Students asked their LLM for an answer right there in class. Ten minutes later, I invited one student to the board to explain the solution using a marker and whiteboard.

The LLM produces an answer but the student must defend it. "Why does this approach work?",  "Where does it fail?" I step in to guide them when they are unable to answer. This is far more fun than just explaining everything myself. It also gives me time to take a break from constantly speaking. One could even argue that this simulates a job interview.

For grading, I used traditional pen-and-paper exams with multiple-choice questions; no LLMs were allowed during the exam. However, I prepared the questions with the help of LLMs, of course!

CV review

From time to time, computer engineering candidates send me their CVs for review. Here are my common responses:

First of all, a CV should not contain any typos; it should show that you are conscientious. CVs usually include a list of projects. For the projects you have worked on, there should be a corresponding GitHub repository. Your GitHub account name should be your real name so that it is easy to tell that it is yours. The repository should not be a random code dump copied from elsewhere or a collection of half-baked code snippets. It should include a decent README and quality code.

In the age of AI, this is fairly easy to achieve, assuming you were highly involved in the project. If I see a GitHub repository without a README, I do not bother looking at the code. It can even work against you, as it may feel like an attempt to pass off unfinished or unclear work. It would also be beneficial to record short videos explaining your projects and upload them to YouTube. I rarely see candidates do this, so it would be a strong differentiating factor.

A well-documented project also shows that you would be easy to work with because you can operate independently. You would not require much hand-holding; when stuck, you would provide timely feedback. Your manager would not have to constantly check on you and would only occasionally need to set priorities and provide direction.

I am also often asked whether certifications are useful. For me personally, they do not mean much. However, if the companies you plan to apply to list certifications in their job descriptions, they can be important for securing an interview. Otherwise, just focus on improving your CGPA and work on projects.

Web/Mobile vs Embedded

Tech evolves at different speeds. In web/mobile development, the landscape resets every few years; React today, something else tomorrow. In embedded systems, the core stays steady for decades. C, RTOS, and hardware fundamentals still rule. Which one should you choose as a career?

Web/mobile shapes how billions interact with information. Embedded shapes how machines sense, move, and endure in the physical world. Both matter, just in different dimensions.

Web/mobile is a newcomers’ arena because it has more job openings than embedded and fast-moving frameworks level the field. Even seniors must relearn constantly, so a quick learner can catch up fast. You can ship real apps in days, no hardware needed. Creativity and adaptability beat deep expertise.

Embedded is a veterans’ stronghold since change is slow but depth matters. Hardware control, timing, and debugging take years to master, and once learned, rarely become obsolete. Senior engineers grow more valuable with each project; their experience compounds.

Web favors fast learners. Embedded rewards deep learners. In web, the new replaces the old. In embedded, the old masters the new.

Curious, fast, and drawn to constant change? You’ll thrive in web and mobile development. Thoughtful, precise, and fascinated by what happens beneath the surface? Embedded systems are your natural habitat.

You could also start with web (especially backend) to learn programming fast (Python or JavaScript). Then move into IoT projects, a perfect bridge between web and embedded. Gradually dive into C/C++, RTOS, and hardware-level concepts.

Work permit in Türkiye for Foreign Students

According to the Turkish Work Permit Evaluation Criteria, a full time foreign software engineer must be paid a gross salary of at least four times the gross minimum wage in order to obtain a work permit in Türkiye. See section A.3 in https://www.csgb.gov.tr/uigm/en/calisma-izni/calisma-izni-degerlendirme-kriterleri. Keep that in mind when you apply for a full time job and they ask you for your salary expectation.

AI Without the PhD: A Developer’s Guide

The AI landscape is vast, but you don’t need a PhD in machine learning to make it work for you. By focusing on practical skills, you can build smart apps without getting lost in the math.

Three levels of understanding:

1. Integration Skills: Know how to call an API that provides AI functionality. Example: Converting speech to commands.
2. Conceptual Understanding: Have a general sense of what training an AI model involves. You don’t need to master neural network mathematics, but you should understand the workflow: Collect data → Split into training/testing sets → Train the model → Evaluate → Deploy.
3. Tool Awareness: Recognize that you can train simple models yourself using online tools (e.g. Edge Impulse) or Python libraries (e.g. PyTorch).

A Practical Example: ESP32 Voice Control with AI (No Cloud Needed!)

1. Record voice samples like “light on” or “light off” as training data
2. To expand the dataset, generate synthetic data automatically with OpenAI's Whisper.
3. Split dataset into 80% training and 20% testing.
4. Train the model with a tool.
5. Test accuracy and retrain if needed.
6. Deploy it to the ESP32 microcontroller.

That’s it. You don’t need to know the inner workings of gradient descent or backpropagation to build something that works in the real world.

And most importantly: create 90% of your code with Claude Code.

Music: Sia - Cheap Thrills

Studying Computer Engineering in Germany: Universität vs Hochschule

If you’re considering studying in Germany, one of the first choices you’ll face is between a Universität and a Hochschule (Fachhochschule / University of Applied Sciences). Both award a  Bachelor of Science (BSc), but they serve different types of students and career goals.

Universitäten focus on theoretical knowledge and research. Professors are evaluated mainly on publications, grants, and PhD supervision, so teaching ability is less emphasized. Class sizes can reach into the hundreds, leaving students with little chance of direct contact with professors. Students dive deep into mathematics, algorithms, and computer science theory, and are largely expected to learn on their own — which can add extra stress for foreign students. This is the ideal path if you want to pursue an academic career.

Hochschulen are industry-oriented. Professors are hired for their industry experience and teaching ability rather than publications. Programs include a mandatory internship semester (Praxissemester) and project-based courses. Class sizes are smaller (often just a couple dozen students), giving students much easier access to professors and helping them graduate with strong connections to employers. This path is perfect if your goal is to graduate as a job-ready engineer.

The most unique feature of DHBW (and all duale Studiengänge in Germany) is that 1 year in advance, you apply first to a partner company (e.g., Mercedes-Benz, Bosch, SAP, Porsche, IBM). The company runs its own selection process. Only if you get a training contract (Ausbildungsvertrag) with the company are you then admitted to DHBW. This is why the probability of graduating in 3 years is high — companies don’t hire people they expect to fail. If you have graduated from a Turkish High School without an IB diploma programme, to apply to the company you first have to either study 1 year in a Turkish university or 1 year in a German StudienKolleg. Since companies take applications for the next year, this would mean losing 2 years. Having an IB diploma saves 2 years.

The dropout rate is around 33% at Universitäten, compared to about 23% at Fachhochschulen, indicating that studying at a Fachhochschule is generally easier.

After graduation, to get a job as a foreigner, the salary offer has to be minimum €43,759.80 yearly gross (as of 2025). This is roughly 1.6 times the minimum wage in Germany.

Türkiye does not have a direct equivalent of the German Hochschule. The closest would be Meslek Yüksekokulu (2-year vocational schools), but those don’t lead to a BSc. All proper engineering BSc programs are run by universities which follow a more research-oriented academic culture and place little emphasis on teaching quality or industry needs. Due to incentives tied to publishing papers, teaching is often viewed as a burden rather than a priority.

Turkish graduates may lack structured internship/practical semesters. To be hired in the tech industry, students must take initiative, actively develop practical skills beyond coursework and demonstrate hands-on experience through internships, freelance work, or GitHub contributions.

AI: More Luck Than Science?

In AI, a lot of progress still comes down to trial and error — and sometimes, plain old luck. We can’t even predict how many images you might need to train a cat classifier to 95% accuracy.

When researchers train giant neural networks, the outcome can swing wildly depending on small, random details. Change the initialization seed, shuffle the data differently, or even let the GPU run in a slightly different order, and you might end up with a model that either crushes benchmarks… or flops.

Big labs try to beat this randomness by brute force — running thousands of experiments in parallel until something works. Smaller teams don’t have that luxury, which is why AI breakthroughs often come from places with deep pockets.

Scaling laws, optimization tricks, and theory give us islands of predictability. But we don’t yet have the “physics of deep learning” — the equations that would let us design a network and know it’ll hit 95% accuracy without a thousand failed runs.

Until then, success in AI will keep feeling less like engineering and more like informed gambling with increasingly sophisticated strategies.