I applied through the company's career site, and a recruiter reached out to me a few weeks later. The first round was with the recruiter and included logistical, behavioral, and situational questions. The next round was with the hiring manager (VP of AI). We discussed my previous projects and experience. He seemed like a very nice and humble person. However, I felt he was super excited about recent advancements in AI and LLMs. My impression was that he was more tactical than strategic in his approach. He was eager to experiment with new AI applications in healthcare without doing any ROI analysis, which gave me a somewhat FOMO-driven impression. The following stage was a take-home assignment. The task involved hosting a small language model and fine-tuning it using LoRA or QLoRA adapters on health data. They said it should take no more than 4 hours, but it's definitely taking longer. The assignment itself was interesting, and I enjoyed working on it. However, I had to pay for cloud resources and GPU usage out of my own pocket. Companies should be more considerate when designing assignments that require model fine-tuning; they should also provide resources to the candidate so that they do not pay for resources out of their own pockets.