Rohit Kulshreshtha (@rohitkulsh) profile picture
Joined October 2009
Photo by Rohit Kulshreshtha (@rohitkulsh)Photo by Rohit Kulshreshtha (@rohitkulsh)Photo by Rohit Kulshreshtha (@rohitkulsh)Photo by Rohit Kulshreshtha (@rohitkulsh)Photo by Rohit Kulshreshtha (@rohitkulsh)Photo by Rohit Kulshreshtha (@rohitkulsh)Photo by Rohit Kulshreshtha (@rohitkulsh)Photo by Rohit Kulshreshtha (@rohitkulsh)Photo by Rohit Kulshreshtha (@rohitkulsh)
Rohit Kulshreshtha retweeted
Had a ton of fun with @CMUDB and @andy_pavlo talking about databases AND AI/robots! SPOILER: we’re giving away our simple recipe for how to work with millions of video hours with just good ol’ parquet, iceberg and mp4. youtu.be/S6_HFMpz_Yk?si=RoME… Now go oneshot this w Fable 🤣
4
2
1
17
1,487
Rohit Kulshreshtha retweeted
Data is always at the core
1
6
256
Rohit Kulshreshtha retweeted
Replying to @lukas_m_ziegler
@eventualai for physical ai data needs
1
1
2
175
Rohit Kulshreshtha retweeted
Fixed it
1
3
5
471
संकट से हनुमान छुड़ावै। मन क्रम बचन ध्यान जो लावै॥ Those who remember Shri Hanuman in thought, words and deeds with Sincerity and Faith, are rescued from all crises in life. My man lived it in his deeds: one last push to open the door.
Captain Smit Machchhar tells PM Modi that he recited Hanuman Chalisa during the Fly Dubai mid-air cockpit emergency. "whenever in difficulty, he recites Hanuman Chalisa", Smit says
53
"We bring you here to a House for the Feeble-minded and the man who won't accept that is the man we want. It's a method that can be cruel but it works. It won't do to say to a man, "You can create. Do so." It is much safer to wait for a man to say, "I can create, and I will do so whether you wish it or not.‟ There are ten thousand men like you, George, who support the advancing technology of fifteen hundred worlds. We can't allow ourselves to miss one recruit to that number or waste our efforts on one member who doesn't measure up."
70
Any problem in computer science can be solved with another level of indirection. If you apply it many times - you have a database. If you see someone applying it too many or too few times, you have the opportunity for a new category of a database product.
1
1
77
Rohit Kulshreshtha retweeted
Every Washingtonian should visit claimyourcash.org to see if you have funds waiting to be claimed. Last year, our Department of Revenue returned $182 million. On Unclaimed Property Day today, we want to get your money back into your pocket.
159
44
5
288
63,823
Having a clickity-clakity feedback for token usage would be pretty awesome.
91
Clear intent helps teams work better. Many past “communication issues” were really gaps in shared intent. Working with AI is making that distinction harder to ignore.
1
1
54
Rohit Kulshreshtha retweeted
Processing 24 trillion tokens without a single crash. @essential_ai built Essential-Web v1.0 using Daft: • 23.6B LLM queries in 7 days • 90K AMD MI300X GPU hours • 32K sustained req/s per VM Scaling vLLM inference at web scale 👇
1
2
11
398
Today we're announcing that Eventual has raised $30M in Seed and Series A funding from @CRV and @felicis as well as @ycombinator, @M12vc and @Citi and others. The AI era needs data infrastructure built for AI, not retrofitted. 🧵
46
40
36
167
61,165
Big moves coming for AI + multimodal workloads, check out Daft: pip install daft
Wrapping up #LaunchWeek with Daft Roadmap! Hear from @Sammy_Sidhu, CEO & Co-Founder of Eventual, about what’s next for Daft and what we’re excited about.
2
130
Daft just made it that much easier to connect real-world, messy data to powerful distributed pipelines. PDFs, videos, embeddings—bring it on.
Introducing User-Defined Data Sources & Sinks. Now you can write to any format – propriety, vectorDB, whatever — with full distributed power in Daft. We even wrote a @trychroma sink in ~100 lines, LIVE demo + PR open 🔥
89
This is a GAME CHANGER for teams juggling legacy + multimodal workloads. Just 2 lines of code to ditch the JVM and run Spark code on Daft !!
✅No JVMs ✅No JARs ✅Local or Distributed ✅One engine for all Daft #LaunchWeek Day 3: SPARK CONNECT FOR DAFT 🚀 Switch from PySpark to Daft with just TWO lines of code and run the SAME Spark queries, but faster and simpler. And easily scale from local to distributed with Ray. No need to maintain separate stacks for both legacy AND multimodal workloads. Same Spark code, radically better experience. 💜 Try it out yourself: pip install “daft[spark]” Learn more about Spark Connect in our docs: getdaft.io/projects/docs/en/…
3
2
66
If you’ve ever OOM’d a data pipeline trying to download and resize images — this one’s for you.
🔥 Fixed batch sizes are old news. DAY 2 OF DAFT #LAUNCHWEEK: Introducing Dynamic Execution for Multimodal Data Processing Daft is built to adapt in real time to multimodal workloads. Resize images, upload to S3, write to parquet – All optimized. All in one pipeline. ✨ Ditch fixed batches and unlock adaptive data flows: pip install daft Check out our docs on dynamic execution: docs.getdaft.io/en/stable/co…
1
40
Rohit Kulshreshtha retweeted
🔥 Fixed batch sizes are old news. DAY 2 OF DAFT #LAUNCHWEEK: Introducing Dynamic Execution for Multimodal Data Processing Daft is built to adapt in real time to multimodal workloads. Resize images, upload to S3, write to parquet – All optimized. All in one pipeline. ✨ Ditch fixed batches and unlock adaptive data flows: pip install daft Check out our docs on dynamic execution: docs.getdaft.io/en/stable/co…
1
4
3
15
1,375
Rohit Kulshreshtha retweeted
🚨 LAUNCH WEEK IS HERE 🚨 Watch Sammy, Co-Founder & CEO of Eventual, kick off an exciting week. We’ve rebranded. Rebuilt. Reimagined Daft. getdaft.io 🧠 Faster docs ⚡️ Cleaner UX 📦 Ready for AI workloads at scale Try out Daft today: pip install daft
2
9
3
18
5,618