I'm Luyi Lin(林路易). You can also call me "speedlast".

Generally, I'm working in AI, struggling to be a care-free engineer and a good-taste researcher.

I care about definitions, real-world problems, solid work and the synergy between system engineering and algorithmic innovation. Recently, my focus has been around:

• Foundation Models: I'm extremely interested in exploring the architecture design and training algorithm of foundation models, validating scaling laws and pushing the boundaries of AI, however small the contribution is.
• AI Infrastructure & Hardware-aware algorithms: I believe that pure system work risks "labor stacking" without algorithmic insights, while pure algorithm work risks "disfalsification" in large-scale scenarios without hardware efficiency.
• AI for math: Trying to solve open questions that have puzzled mathematicians for decades or even centuries could be fun.

Beyond research and work, I cherish individual independence and dignity, traversing existential struggles with a lifelong quest for love and freedom.

On Loyalty: Ride or Die. Through thick and thin.

就此拜过昭烈武侯。