Presenting our Embedding-to-Prefix work at #NeurIPS2025, projecting user embeddings into LLMs at inference time for personalization. Excited to discuss with folks working on scalable preference learning, RLHF, and personalized LLMs/agents.
1
4
308
the bottleneck in scaling agentic systems isn’t the optimizer, it’s execution. once you can actually generate + resolve diverse candidates, calibrated rewards + preference optimization become effective. feels like a general pattern for agentic ai.
1
2
470
new spotify research blogpost where we dive into this: personalizing agentic ai to users’ musical tastes with scalable preference optimization: research.atspotify.com/2025/…
1
150
I successfully defended my PhD! Thanks to @kgajos, @pmphlt, and @danmcduff. I could't have imagined a better committee! 😃🤓🤩
8
50

