Prasang Upadhyaya (@prup) header
Prasang Upadhyaya (@prup) profile picture

AI/ML @dropbox. Past: @twitter, @uwcse, @iitbombay. Lesser half of @ketkisharma.

San Francisco
Joined May 2009
Photo by Prasang Upadhyaya (@prup)Photo by Prasang Upadhyaya (@prup)Photo by Prasang Upadhyaya (@prup)Photo by Prasang Upadhyaya (@prup)Photo by Prasang Upadhyaya (@prup)
Prasang Upadhyaya retweeted
Incredible case study that's readable yet filled with practical approaches and tactics: ✔ Use GEPA to get near-frontier performance at 1/100th the cost. ✔ Quickly evaluate new models, without having to rewrite prompts. ✔ Evolve & maintain your programs w/o full rewrites
How we used DSPy to turn our relevance judge into a measurable optimization loop, making it more reliable and scalable in Dropbox Dash.

6
17
1
219
29,222
Prasang Upadhyaya retweeted
💡 Two arXiv papers published in recent days (one from us, one from TUD) reach the same conclusion: LLMs can now generate C++ code for SQL processing that outperforms classical database systems. ⚙️ Our code generator is based on Claude Code and exploits multiple agents working in parallel. Each agent performs tasks typically associated with different components in a #DBMS, such as workload analysis, query optimization, or physical design tuning. 📊 We compare to various classical #DBMS such as DuckDB, ClickHouse, Umbra, MonetDB, and PostgreSQL, finding that the agent-generated code is often significantly faster. Code generation costs are moderate (<$20), making the approach practical for frequently executed queries. 🤖 Analyzing generated code, we find that agents exploit various optimization techniques, including query-specific data structures, as well as low-level optimizations that are specific to the hardware cache hierarchy of our server. 📃 Paper: arxiv.org/pdf/2603.02081 💾 Code: github.com/SolidLao/GenDB 🌐 Site: solidlao.github.io/GenDB @lojil192574 #LLM #Databases #AI #DB
3
7
1
31
3,543
Prasang Upadhyaya retweeted
We just rolled out our Spring ’25 release with some big updates to Dropbox Dash, our AI search and knowledge management product. Here’s what’s new: 📷 Search across video, audio, and images — no more digging for cryptic file names. You can even use Dash to search for people you work with. (1/2)
12
11
9
84
38,703
Prasang Upadhyaya retweeted
We describe a method for maximising long term engagements by using model-based reinforcement learning. The policy is then used to make decisions about whether to send push notifications or not. More details are in the paper if you are interested. arxiv.org/abs/2202.08812
1
5
11
Prasang Upadhyaya retweeted
Please help spread the word. I'm looking for strong PhD candidates and a postdoc to work with me at @UofTCompSci. If you're excited to work in algorithms, optimization, theoretical machine learning, come join us.
6
125
4
253
Prasang Upadhyaya retweeted
My team is hiring a Staff PM to drive the Home tab and feed experience on @Twitter. Massive scope and impact. Job post not up yet. DM if you're interested or know someone great.
5
38
12
213
Prasang Upadhyaya retweeted
Drumroll please! Our first-ever algorithmic bias bounty at #defcon wrapped this weekend, and the winners were announced at yesterday’s #defcon @aivillage_dc workshop. ICYMI, you can watch a recording of the presentation here youtu.be/KLwdzLiJywI
3
44
33
127
Prasang Upadhyaya retweeted
We are looking for a mobile engineer with react native experience to join our team at @BicycleHealth. DM me if you are interested, and pls tag anyone who might be interested! #hiring #softwareengineering #reactnative #HealthTech
1
1
6
Prasang Upadhyaya retweeted
So impressed by this work by @RyanMarcus @tim_kraska @parimarjan @hongzimao @tatbul and Mohammad Alizadeh on learned optimizers. One of my favorite pieces of recent research in the DB community. Congrats on the @SIGMOD_2021 Best Paper Award! learnedsystems.mit.edu/bao-d…
1
6
1
51
Congratulations to Yossi Azar, Andrei Broder, Anna Karlin, Michael Mitzenmacher, and Eli Upfal for receiving the 2020 ACM Paris Kanellakis Theory and Practice Award. Their Balanced Allocations framework has had a wide-spread practical impact. https://xitter.cf/t.co/8dy04XpGn3
37
8
141
Prasang Upadhyaya retweeted
Just in: The U.S. Supreme Court, in a 6-2 opinion, found that Google’s use of Oracle's lines of code when creating its Android operating system constituted “fair use” under the copyright laws.
35
295
107
1,093
Columbia's Al Aho & Stanford's Jeffrey Ullman receive the 2020 #ACMTuringAward for fundamental algorithms and theory underlying programming language implementation & for synthesizing these results in their highly influential books. https://xitter.cf/t.co/wApf2h4W4Q @CUSEAS @StanfordEng
15
299
110
708
Prasang Upadhyaya retweeted
After a long wait, we are finally announcing the #Recsys2021 challenge! This year we are releasing around 1 billion samples, largest social network dataset by far, with an added focus on fair recommendations. Please checkout our website recsys-twitter.com/ for more details.
We are proud to announce the #RecSys2021 challenge! The data for this year's challenge is provided by @TwitterResearch. This year's goal is two-fold: predict different engagement types, while providing fair recommendations. More details on our website: recsys.acm.org/recsys21/chal…
Prasang Upadhyaya retweeted
This week I started a new role at @Twitter, leading a team to work on what I'll call “social science for machine learning.” WERE HIRING. We're tackling questions like...
43
182
35
1,095
If you're interested in DB internals, stop what you're doing and watch @CMUDB Quarantine Talk from Nico+Cesar about @SQLServer's Cascades query optimizer: youtube.com/watch?v=pQe1LQJi… Many talks this semester were good. This one is the best. My thread provides key takeaways
4
49
7
285
Prasang Upadhyaya retweeted
Snippet 1 from the #NeurIPS2020 tutorial: @balajiln What do we mean by uncertainty and out-of-distribution robustness? nips.cc/virtual/2020/tutoria…
3
70
11
269
Prasang Upadhyaya retweeted
New blog post: Neural Networks (Maybe) Evolved to Make ADAM the Best Optimizer For the first time, I wrote a blog post without math :) I discuss a *conjecture* I have regarding Adam and the way the deep learning community produces new ideas parameterfree.com/2020/12/06…
14
129
34
808