First 45-60m interview with hiring manager. Pretty simple, nothing too technical.

The guy asked if I have worked with video before and I clearly answered "no". He then asked if I would be interested to learn and I said "definitely yes".

The HR also emphasized that the role is about data modelling, video annotation and cataloguing using existing models and not training or custom video processing algorithms.

Nevertheless, I got completely the opposite feeling from the feedback on the technical exercise.

Although I solved the asked problem, the feedback was that they were expecting me to go deeper into the problem space, solving cases that the off-the-shelf scene detector will fail as well as more debugging and analysis on the detector results. If this is the case, then don't waste people's time.

There was also a 2nd part to the exercise: design a video ingestion and annotation pipeline. The instructions were very high-level—describe the components, inputs/outputs, and tools used. Again, they were expecting to see something far more detailed than asked.

Design and implement a scene detector