I moved from network training to deployment optimization and am starting to think that was a right decision [D]
retained text
TL;DR: this is just a thought dump about how I moved from network training to deployment and why I am making peace with (maybe) not getting back to model training, nothing too interesting. I used to work at a reasonably good startup developing and training models. In 2024, I got an offer from FAANG but the catch was that the team worked on the deployment of networks. I have a PhD and I had always trained models but I was never the top performer in the team frankly (I was still top 20%, those top guys are at Nvidia and Tesla AI now, just for context). I think there were others who had more novel ideas than me, read more papers. If I tried 10 ideas, 1 would work and from outside it looked like more of their ideas worked. Since I wanted to leave the startup anyway and with no other offer, I decided to join FAANG. I genuinely thought that was a downgrade. Then LLMs started getting so competent. I think I made the right decision, let me explain. I have been trying to learn diffusion models (I have never worked with them) and these frontier models are explaining in such a good manner, able to generate visualizations to back up their explanations. I feel like I would stand no chance if I went back into model training. They have better ideas than I do. Unless I work on GPT-7 (which I cannot), I feel like these models are plenty intelligent to propose ideas. Deployment is a different world. Most edge architectures are confidential and heavily protected. An LLM won’t have any idea out of the box and even with context, a lot of knowledge is empirical and tribal. I still have FOMO about model training but slowly making peace with it that getting back to it may not be worth it.
document fetched 2026-10-20 00:27 · canonical https://reddit.com/r/MachineLearning/comments/1x9k4si/i_moved_from_network_training_to_deployment