Shape your SOP
around the lab you chose.

Research Statement

My interest in federated learning for privacy-sensitive applications emerged from a challenge I encountered during my final-year project: training a diagnostic model across three hospitals without centralising patient data. The constraints I faced (heterogeneous data distributions, strict audit requirements, and latency-sensitive inference) led me to the literature on cross-silo FL, where Prof. Chen’s NeurIPS 2024 work on differential privacy under non-IID splits directly addressed the gap I had struggled to solve.

I want to extend this work toward real-time inference in cross-institutional settings specifically the latency versus privacy trade-off when ε-differential guarantees must hold under adversarial participation. This aligns with the Chen Lab’s stated direction toward deployable, privacy-preserving infrastructure for clinical AI.

[Draft continues: Academic background · Research methodology · Why MIT]

Highlighted text shows where the draft connects to Prof. Chen’s specific research. Other targets would need different papers and a different research-fit angle.