Describe A challenging technological problem you faced

One of the most daunting technological problems I’ve ever tackled was a crippling latency issue with a real-time facial recognition system I developed for my final-year engineering project, last April in my university’s embedded systems lab. The system was designed to verify campus visitors’ identities within 0.5 seconds, but initial tests revealed a 2.8-second delay — far too slow for practical deployment. What made it even more challenging was that I only had two weeks left before the submission deadline, and my prior knowledge of edge computing was purely theoretical, with zero hands-on experience. I spent three days of fruitless debugging before I almost threw in the towel. Eventually, I stumbled upon a research paper on knowledge distillation for lightweight AI models, so I compressed the pre-trained recognition model by 60% while retaining 98% of its accuracy. I also optimized the frame-sampling logic to skip redundant low-resolution frames. After two all-nighters of tweaking parameters, the latency dropped to just 0.32 seconds, exceeding the project’s requirements. This experience really honed my problem-solving skills and sparked my passion for edge AI innovation.