Deepti Raghavan
@deeptir18
Followers
754
Following
193
Media
5
Statuses
43
Assistant Professor of CS at Brown
Providence, RI
Joined June 2011
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Madison Square Garden
• 807121 Tweets
Arsenal
• 344596 Tweets
#MexicoGP
• 192321 Tweets
Puerto Rico
• 153886 Tweets
Perez
• 133113 Tweets
Jets
• 125970 Tweets
Boulos
• 121513 Tweets
Eagles
• 120733 Tweets
Sainz
• 119722 Tweets
Charles
• 116759 Tweets
Ferrari
• 100048 Tweets
Verstappen
• 92564 Tweets
Norris
• 90547 Tweets
Checo
• 87076 Tweets
Browns
• 84868 Tweets
Fortaleza
• 84755 Tweets
Nunes
• 79949 Tweets
#OMPSG
• 77463 Tweets
Tarcísio
• 76233 Tweets
Bills
• 74048 Tweets
Franco
• 71337 Tweets
Ravens
• 70935 Tweets
Leclerc
• 55327 Tweets
Lawson
• 52666 Tweets
Jameis
• 50306 Tweets
Lando
• 49794 Tweets
Bengals
• 49342 Tweets
Bears
• 47968 Tweets
Hamilton
• 40699 Tweets
André Fernandes
• 37657 Tweets
Raiders
• 34737 Tweets
Colts
• 29607 Tweets
Tony Hinchcliffe
• 29087 Tweets
Rodgers
• 25791 Tweets
Caleb
• 24810 Tweets
Porto Alegre
• 24605 Tweets
Dr. Phil
• 23208 Tweets
Seahawks
• 21897 Tweets
#GHDBT8
• 19102 Tweets
Hulk Hogan
• 15197 Tweets
#MexicanGP
• 11089 Tweets
I'm very excited to be starting at
@BrownCSDept
this fall!
Please welcome Deepti Raghavan, joining us this fall as assistant professor! A recent graduate of
@Stanford
, her research interests focus on operating systems, networking, and machine learning systems:
1
6
82
25
20
426
Stanford CS is running an application support program for underrepresented students. If you’re considering applying to the Computer Science PhD program at Stanford, we'll do our best to give one round of feedback on your application. Apply by October 29:
4
127
359
We’re excited to share our ongoing work on ALTO, a network orchestrator for efficiently serving compound AI systems. ALTO can improve serving throughput for a complex chatbot verification pipeline by 3x while reducing tail latency by 1.8x.
1
11
49
I will be recruiting PhD students this upcoming cycle, to start in fall of 2025. Please apply if you are interested in operating systems, networking or machine learning systems.
2
0
18
I am extremely grateful to my family, friends, collaborators and mentors --
@matei_zaharia
, Phil Levis, and
@schemeprincess
-- for helping me get to this point. Thank you for making my PhD both fun and rewarding!
2
0
8
Generative language models emit tokens incrementally, which is especially useful in the context of compound AI systems because we can stream partial outputs between pipeline stages to overlap computation.
2
0
6
Many compound AI systems combine generative LMs with external tools like retrievers in a pipeline to solve challenging tasks. This work studies how to effectively distribute and parallelize these pipelines at scale.
1
0
5
This is joint work with Keshav, Shahir, Thejas, Neha, Pratiksha,
@matei_zaharia
, and Phil. Check out our paper here:
1
0
4
Some stages are stateful, meaning they have to aggregate partial outputs across a stream. For such cases we need to route all stream data through a consistent stage instance to ensure correct aggregation. We call this aggregation-aware routing.
1
0
4
We discuss some preliminary ideas on designing an interface to enable localized aggregation-aware routing and a distributed prompt-aware scheduling algorithm in our paper.
1
0
3
Nice work
@dmshanmugam
!!!!!!!
Machine learning model uses your heartbeat to assess risk of cardiovascular death:
1
38
93
0
0
2
Two new challenges emerge when we stream partial outputs between pipeline stages: correctness and load balancing.
1
0
3
Each LM stage can produce dynamic fan-out based on how many partial outputs it emits / how frequently the stage is invoked. To handle this fan-out and efficiently load balance across stages we need distributed prompt-aware scheduling.
1
0
3
We are actively looking for more applications to serve on top of ALTO. If you’re working on compound AI systems and want to serve them at scale, please reach out!
1
0
2
@jaewon_chung_cs
@matei_zaharia
The streaming is when downstream tasks require some part of the LM output from the previous stage (not always at the granularity of a token). Claim extraction streams at the granularity of individual sentences to the query generation state, which generates a query per sentence.
1
0
2
@kellycochra
@otherskal
@andrewkuznet
@LizIzhikevich
It should be open soon! We will post more details then.
0
0
1