Feedback from people who have completed the programmes

I have tried two self-paced ML courses before this one. Both times I ran into the same problem: I could follow the notebook but did not understand why any of it worked, and there was nobody to ask. The ten-week programme here is different mainly because the instructor actually looks at your assignment code and writes back. The feedback on my week-five submission pointed out a leakage problem I had missed and explained why it would have made my evaluation numbers meaningless. That kind of note is not something an auto-grader can produce.

ML Foundations · June 2025

The short course was good value for three weeks. The workshop on handling missing values was particularly useful — the instructor drew a clear line between imputation that is safe to do and imputation that leaks information from the future, which is not something I had seen explained well before. I did find the DVC material in week three moved quite fast. I would have preferred an extra session on it, but I understand why they had to keep to the three-week scope. I took notes and could fill in the gaps later.

Data Preparation · May 2025

I joined the LM Engineering programme because my team had started using an LLM API and I wanted to understand what we were actually doing, not just calling the API and hoping. The section on offline evaluation harnesses in weeks nine and ten was worth the twelve-week commitment on its own. We had been relying on vibes to assess whether a prompt change was better or worse. After the programme I set up an evaluation script that gave us actual numbers. The running project format also forced me to actually build something rather than just follow along with examples.

LM Engineering · April 2025

I was sceptical that I could keep up given my Python is not strong. I did meet the prerequisite — I can write functions and set up environments — but I expected to struggle with the maths. The linear algebra refresher in week one was one of the better learning experiences I have had. It was just enough without being a full university module. By week four I understood what a loss function was doing in a way that actually meant something. The Monday/Wednesday evening format worked well with my schedule.

ML Foundations · May 2025

The running project structure is the right call for this kind of material. LLM application development has too many moving parts to learn through standalone exercises. Building something continuously over twelve weeks meant the decisions I made in week three had consequences in week nine, which is actually how real projects work. One thing I would change is more time on cost and latency budgeting in week six — the section felt rushed. But overall I came away with a working internal tool and a much better understanding of where RAG pipelines actually fail.

LM Engineering · June 2025

I took the Data Preparation short course specifically to decide whether I was ready for the ten-week ML programme. That is exactly the use case they describe, and it worked. Three weeks is enough time to find out whether you can keep up with the pace and whether the material is at the right level for where you are. I am now enrolled in the October ML Foundations cohort. The short course also introduced me to DuckDB, which I had not used before, and I have been using it at work since finishing.

Data Preparation · June 2025