My time spent as a Research Engineer at prestigious institutions like ENS Ulm and École Polytechnique has been a formative experience in my journey into AI. It's one thing to learn about algorithms in a classroom, but it's another entirely to implement and extend state-of-the-art frameworks like CoVR for composed video retrieval. Working on projects such as 'Improving Vision Language Models' taught me the grit required in research. A significant portion of the work isn't about glamorous breakthroughs, but about rigorous, systematic experimentation. Managing complex configurations with Hydra, handling terabytes of data on GCP, and meticulously evaluating results against benchmarks—these are the real activities that push the needle forward. My biggest takeaway is that innovation often lies in the details. Enhancing class discrimination by engineering a new penalty function or exploring non-linear mappings might sound niche, but these small modifications are what lead to more robust and efficient models. It's a testament to the idea that true progress is built incrementally, one carefully tested hypothesis at a time.
Bradfo'siPhone