Desh Raj
2,244
posts
@rdesh26
New York, NY
Joined September 2009
Pinned
I’m happy to share that I’m starting a new position as Senior Research Scientist at
@nvidia! Looking forward to open science for speech full-duplex models :)
After 2 wonderful years, I left Meta this week. During this time, I worked on several projects related to speech and LLMs: - Built the first multi-channel audio foundation model with M-BEST-RQ (arxiv.org/abs/2409.11494) - Made ASR with SpeechLLMs faster (arxiv.org/abs/2409.08148)
At my French class, someone asked me what I do for work, and to keep things simple, I said I did AI research. "AI? Didn't that only just become a thing?"
Me: "What are you working on?" PhD student: "Continual learning" Me: "Cool, just like everyone else" PhD student: "Yeah, but I've been interested in it before it became popular" had this conversation verbatim multiple times this year
Ça fait deux mois que j'attends mon remboursement pour un train annulé
@TGVINOUIça a pas de bon sens ces délais?!
Thoughts while attending the Omnimodal Encoders presentation at JSALT 2026: 1. We don't need modality-wise encoders. It made sense for small backbones (8B to 30B) where the encoder was still a reasonable fraction of total params. But as we move to bigger backbones, it would be
Attending the JSALT 2026 presentations at
@jhuclspand realized I have entered my influencer era now 😆
