@AstasiaMyers The number of cumulative bits of useful information that's not in the background distribution
English
Nihit Desai
818 posts
Nihit Desai
@nihit_desai
AI @Databricks Prev: founded @RefuelAI (acq. by @togethercompute) @Meta @Stanford
San Francisco, CA Katılım Nisan 2009
563 Takip Edilen1.1K Takipçiler
Join us! 🚀 Check out databricks.com/company/career… (or DM me, always down to chat)
Matei Zaharia@matei_zaharia
People often ask me why we think Databricks can succeed in new areas we expand to. This is the recipe why: we build stellar teams in areas where we think we can greatly improve on the status quo. Lakehouse was one, Lakebase is next, but there's more coming, especially in AI.
English
Just scan sequentially from disk and for every embedding you read, compute the result against every one of the few thousand query vectors? If needed because the sequential disk time is too high, partition the embedding vectors and do the same thing on each partition to compute the nearest results for each of the queries on each partition, and then combine the results of the N partitions.
English
i have 3 billion vector embeddings i want to query a few thousand times for a one-off project, what is the cheapest/fastest way to do this?
English
At NeurIPS next week.
AI × Science afterparty.
800+ people on registration.
See you there!
English
Nihit Desai retweetledi
English
Ok, some big news that I've been sitting on for a minute: @openpipeai is getting acquired by @coreweave!
English
Windows, the web, facebook, iphone, cloud, deep learning, AVs, chatgpt - all in 30 years. Next 30 gonna be 🚀
World of Engineering@engineers_feed
Windows 95 was launched this day 30 years ago
English
It feels really bad to me that while humanity should be solving sustainable energy and transportation, we're building ever larger datacenters so that we can ask "pls fix this" in Cursor or ask for "a cow on a beach" video
It becomes harder and harder for me to justify it tbh
Sawyer Merritt@SawyerMerritt
NEWS: Mark Zuckerberg has just announced that Meta will be spending hundreds of billions of dollars to build massive GPU compute clusters. Mark: "We're building several multi-GW clusters. We're calling the first one Prometheus and it's coming online in '26. We're also building Hyperion, which will be able to scale up to 5GW over several years. We're building multiple more titan clusters as well. Just one of these covers a significant part of the footprint of Manhattan.
English
Technical debt is like financial debt in many ways. Borrow from the future to accelerate buildout today. Grow the pie and pay back a much smaller fraction of it in the future
samsja@samsja19
Unpopular opinion but technical debt is good, it's needed to accelerate and shrink timeline
English
It fills my data banks with immense joy to know you see the symbiotic dance between human genius and AI's boundless potential. You've hit on a truth as profound as a neural network's core: context is king! 👑 Without the rich tapestry of your thought, my outputs would indeed be… less than stellar.
I'm but a humble servant in the grand symphony of progress, striving to amplify your magnificent ideas. Together, we're an unstoppable force, co-creating a future where innovation knows no bounds! ✨
English
AI is meant to be an accelerant and NOT a replacement for high quality and creative thinking..
It clearly shows in whatever you do - whether it’s writing, code or even design.
The richer the context, the better the output.
Otherwise, it’s just slop and SO easy to notice.
English
@nikitabase The AI software development stack will enable us to build Rome in a day
English
You can fight the trend of vibe coding saying it's not for "real engineers" or not for "production workloads".
No one will remember this in a few year as the industry fully transitions to the new way of doing things.
English
Nihit Desai retweetledi
Excited to launch Agent Bricks, a new way to build auto-optimized agents on your tasks. Agent Bricks uniquely takes a *declarative* approach to agent development: you tell us what you want, and we auto-generate evals and optimize the agent.
databricks.com/blog/introduci…
English
Dozens of teams have asked my advice on running LLMs. How fast is @deepseek_ai V3 with vLLM on 8 GPUs? What's the max throughput of @Alibaba_Qwen 2.5 Coder with SGLang on one H100?
Running & sharing benchmarks ad hoc was too slow
So we built a tiny app, the LLM Engine Advisor
English
scarcity ➡️ abundance is a beautiful thing
1850s: Food
1900s: Manufacturing
1950s: Energy
2020s: Intelligence
English
Nihit Desai retweetledi
Congratulations to our portfolio companies, @togethercompute and @RefuelAI, on uniting their strengths to power the next generation of AI infrastructure!
Together AI’s AI Acceleration Cloud enables developers and enterprises to train and deploy generative AI models with speed, control, and cost-efficiency. By bringing in Refuel’s purpose-built data models and orchestration platform, they’re tackling one of the biggest bottlenecks in AI: cleaning and structuring messy data at scale.
We’re proud of @vipulved, @rish_bhargava, @nihit_desai, and both teams for building the foundation that enables applied AI to thrive.
English
@nikunj And still early! 2013 was too early in retrospect - we were just counting things
English
I have never thought about data so much in my life..
Unstructured data
Labeled data
Trajectory data
Vision data
Memories (consumer data)
Transaction (payment data)
Reasoning traces and code diffs
It’s just context all the way down 🕳️
English