Ali Ghodsi Profile
Ali Ghodsi

@alighodsi

12,212
Followers
208
Following
3
Media
102
Statuses

Databricks CEO & Co-founder, UC Berkeley Faculty

Berkeley, CA
Joined January 2010
Don't wanna be here? Send us removal request.
Pinned Tweet
@alighodsi
Ali Ghodsi
1 month
Today we released an open source model, DBRX, that beats all previous open source models on the standard benchmarks. The model itself is a Mixture of Experts (MoE), that's roughly twice the brains (132B) but half the cost (36B) of Llama2-70B. Making it both smart and cheap. Since…
44
223
1K
@alighodsi
Ali Ghodsi
1 year
Free Dolly! Introducing the first *commercially viable*, open source, instruction-following LLM. Dolly 2.0 is available for commercial applications without having to pay for API access or sharing data with 3rd parties.
55
448
2K
@alighodsi
Ali Ghodsi
1 year
We are open sourcing Dolly, a ChatGPT-like model that can do instruction following, created for $30, trained 3 hours on 1 server. The secret in magical human-like interactivity probably lies in a small dataset.
40
471
2K
@alighodsi
Ali Ghodsi
6 months
The founders of Databricks put together this strategy blog on where we think data platforms are headed in the future. We're moving Databricks quickly in this direction. This is very exciting and is the outcome of the MosaicML acquisition we did earlier this year!…
14
144
812
@alighodsi
Ali Ghodsi
3 years
Excited that our #Lakehouse paper got published at #CIDR21 : it shares our vision of the Lakehouse: a new type of data platforms that are completely open, have full support for #machinelearning , while supporting all traditional #datawarehouse workloads.
Tweet media one
6
119
301
@alighodsi
Ali Ghodsi
2 months
I think this will mark an important milestone for Gen AI. The spotlight has been on the capabilities of LLMs (scaling laws, leaderboards, etc). But it's now clear that LLM performance alone will be meaningless. You will need a Compound AI system to get the best performance out of…
@matei_zaharia
Matei Zaharia
2 months
Interesting trend in AI: the best results are increasingly obtained by compound systems, not monolithic models. AlphaCode, ChatGPT+, Gemini are examples. In this post, we discuss why this is and emerging research on designing & optimizing such systems.
30
261
1K
6
44
241
@alighodsi
Ali Ghodsi
1 year
Dolly LLM now on @huggingface , check it out! Cost under $100 to produce on a few machines for a few hours!
4
44
230
@alighodsi
Ali Ghodsi
1 year
Love this integration between @huggingface and @databricks . Concretely, you will be able to train your own LLM from using Spark and Transformer/Dataset with this tight integration:
4
45
211
@alighodsi
Ali Ghodsi
1 month
There is so much focus on the standard LLM benchmarks (MMLU, ARC, GSM8k etc), but for enterprises the only thing that matters is how well the AI does on the domain specific tasks. Check out a comparison between DBRX and GPT4 on these domain-specific benchmark datasets.
@JuliaANeagu
Julia Neagu
1 month
@DbrxMosaicAI DBRX outperforms @OpenAI GPT-4 on realistic, domain-specific benchmark datasets. For example, on a customer support summarization use-case👇👇👇 Still neck and neck but it shows that open models can be the no-brainer choice for actual enterprise applications.
Tweet media one
16
70
203
7
35
201
@alighodsi
Ali Ghodsi
1 month
This of awesome. Try it out on perplexity.
@AravSrinivas
Aravind Srinivas
1 month
The world's best open-source chat LLM, DBRX, is now available for free, on . Perplexity Labs Playground basically has everything that you need for chat, for free, with better LLMs (Haiku, DBRX, Sonar) than 3.5-turbo, the model powering free chatGPT. Curious…
Tweet media one
104
146
1K
4
19
198
@alighodsi
Ali Ghodsi
1 year
Very excited to release #MLflow 2.3 with native support for LLMs, integrations with Hugging Face transformers, models calling OpenAI, integrations with LangChain. #LLMOps taking off!
8
37
194
@alighodsi
Ali Ghodsi
3 years
@technology @emilychangtv Why do you have to pick between diversity and merit? Make your work env. inclusive, decrease bias in hiring, as a leader don't make statements that alienate large groups. That'll give you a competitive advantage to a talent pool that won't join unwelcoming companies.
6
35
189
@alighodsi
Ali Ghodsi
10 months
Closing out an incredible @databricks #DataAISummit with a keynote from @pmarca . We agree that AI is a way to make everything we care about better. Marc's article is required reading for anyone thinking about AI. Tune in to hear us discuss >
Tweet media one
4
16
171
@alighodsi
Ali Ghodsi
10 months
Excited to be launch partners with Meta on the Llama2 release. This move by Meta will have a big positive impact on the industry and ecosystem. Technically, the first version of Llama already was available to everyone except anyone who had a commercial use case could not innovate…
4
21
169
@alighodsi
Ali Ghodsi
1 month
Is $20/month for an LLM chatbot too much or too little? It's the perfect price if you're reading 2.5 books a day, i.e. 17 hours continuous reading every day of the year! Here is an oversimplified gross margin calculation. OpenAI offers GPT3.5-turbo API and charge $2 per million…
18
25
159
@alighodsi
Ali Ghodsi
1 year
Excited to launch the first two #LLM MOOC courses with @edXOnline . Learn about prompt engineering, vector embeddings, retrieval, chains, and MLOps. Learn how to create your own LLM from scratch on a data lakehouse!
2
44
141
@alighodsi
Ali Ghodsi
1 year
We are also releasing the first 15k human generated high quality dataset that can teach models to interact like humans (instruction following).
1
13
144
@alighodsi
Ali Ghodsi
3 years
Join me today at 5pm on Clubhouse as I interview @bhorowitz and @pmarca on their journey founding and creating @a16z together.
6
10
122
@alighodsi
Ali Ghodsi
3 years
This 👇
@jay_drainjr
jaydrainjr.eth
3 years
Consumer vs Enterprise: Clubhouse raises at $1bn val and the world can’t stop talking about it Databricks raises at a $28bn val and 90% of us don’t know what lakehouse architecture is
0
16
113
8
10
90
@alighodsi
Ali Ghodsi
3 years
@technology @emilychangtv and... Databricks is hiring 😉
0
7
73
@alighodsi
Ali Ghodsi
3 years
Join @bhorowitz and @pmarca and me today on Clubhouse at 5pm PST to hear about the past, present, future of AI.
2
8
66
@alighodsi
Ali Ghodsi
3 years
Excited for these investments from all our cloud partners as we democratize AI to all enterprises.
@databricks
Databricks
3 years
“The future happens faster than people think. They look back and say, my God, I can’t believe the world we live in now compared to 10 years ago...And the great companies are the ones that bet on those secular trends.” - @alighodsi
1
21
68
9
14
68
@alighodsi
Ali Ghodsi
1 year
Why doesn't someone put together two LLMs in an generative adversarial setting and have them challenge each other to prove all open math conjectures out there?
12
6
66
@alighodsi
Ali Ghodsi
3 years
I’m excited to announce @databricks ’ acquisition of 8080 Labs! This is a strategic foray into the low-code/no-code space that extends the #lakehouse with a direct offering to the citizen data scientist.
1
6
65
@alighodsi
Ali Ghodsi
3 years
Join clubhouse and listen to @bhorowitz and @pmarca and me chat about business strategy on Boss Talk!
3
3
63
@alighodsi
Ali Ghodsi
2 years
The best Data Warehouse is the Lakehouse. Retweeting this Databricks ad, if nothing, because the thread below is a must read :-).
@databricks
Databricks
2 years
Data warehouses can't handle 90% of your data. Lakehouses do what warehouses can't. And everything else — from BI to AI. Discover Lakehouse.
12
12
82
3
11
58
@alighodsi
Ali Ghodsi
2 years
This picture is why companies struggle with AI.
@pwendell
Patrick Wendell
2 years
Unfortunately for all of us, this is a highly accurate depiction of #MLOps . @Databricks is focused on making this much simpler. From: "Machine Learning Operations (MLOps): Overview, Definition, and Architecture". Dominik Kreuzberger, et al
Tweet media one
57
828
4K
4
5
57
@alighodsi
Ali Ghodsi
3 years
Tune in. It’ll be fun!
@bhorowitz
benahorowitz.eth
3 years
I'm discussing boss stuff Tuesday, Feb 9 at 5:00 PM PST with @alighodsi , ceo of Databricks. @joinclubhouse . Join us!
31
34
273
4
14
57
@alighodsi
Ali Ghodsi
1 year
Very excited to have the Okera team part of Databricks!
@steph_palazzolo
Stephanie Palazzolo
1 year
Exclusive: @databricks is acquiring data security startup @okerainc as customers push to safely use their proprietary data in custom LLMs. I chatted with CEO @alighodsi on how the deal came together and why governance is their #1 priority right now.
2
13
26
4
9
57
@alighodsi
Ali Ghodsi
3 years
#DataAISummit registration is open! Join this free event & experience an incredible lineup of speakers, 200+ sessions & more. I’m especially excited for our keynotes, including the first U.S. Chief Data Scientist @dpatil
0
19
53
@alighodsi
Ali Ghodsi
3 years
@martin_casado (1/2) I find that there is a more common problem. Great product, bottom up motion working really well. But to really grow to massive revenue you will need enterprise GTM (even Tableau, Salesforce, etc had to eventually do this). But startup product founder resists this change...
3
4
48
@alighodsi
Ali Ghodsi
3 years
Hear us on Clubhouse!
@bhorowitz
benahorowitz.eth
3 years
I'm discussing “Boss Talk” with @alighodsi , Felicia Horowitz, and @pmarca , Today, Feb 9 at 5:00 PM PST on @joinclubhouse . Join us!
13
14
134
5
1
47
@alighodsi
Ali Ghodsi
10 months
We're going to democratize AI and make the Lakehouse the best place to build generative AI and LLMs. We'll talk more about @MosaicML at this week's #DataAISummit . The conference is sold out, but you can still tune in virtually!
1
3
47
@alighodsi
Ali Ghodsi
10 months
When the deal closes, @NaveenGRao @hanlintang @jefrankle and the INCREDIBLE @mosaicML team will join, and we'll give customers the ability to train their own models, with their own data, for their use cases. Together, we’ll do what couldn’t be possible alone.
6
2
43
@alighodsi
Ali Ghodsi
1 month
It keeps getting smarter, we look at errors and are going to have the “diagnose error” functionality everywhere. Why is my job failing? Diagnose Error. Why does this dashboard not render? Diagnose error…
@JohannesVink
Johannes Vink
1 month
I hit an error in my notebook and the @databricks Assistant politely told me what the cause was, how to work-a-around it, but it also told me that if I used another function, I would not have to use the work-a-round. Wow. (that compensated for all the times it was plain wrong!)
4
3
40
4
3
37
@alighodsi
Ali Ghodsi
10 months
Since November of last year, every customer I meet with asks me, “How do I train and tune my own models? How do I keep my own data and own the IP?” When we asked around about the companies at the forefront of this, everyone said @MosaicML .
2
0
35
@alighodsi
Ali Ghodsi
3 years
Join us tonight on ClubHouse at 830pm PDT!
@BestLiveAudio
The Best of Live Audio
3 years
Which company is backed by the four cloud titans—Amazon, Google, Microsoft and Salesforce—in addition to investors like a16z? Join us Wednesday! @databricks @alighodsi @nrmehta @ashugarg @soccolich #cloud #data #AI
Tweet media one
3
14
20
2
10
33
@alighodsi
Ali Ghodsi
1 year
Don't miss Data + AI Summit this year!
@matei_zaharia
Matei Zaharia
1 year
#DataAISummit Call for Presentations ends Jan 13! Submit your case studies, research or best practices around #Lakehouse , MLOps, streaming, and data. Last year's event included 60,000 data experts.
3
42
74
1
6
33
@alighodsi
Ali Ghodsi
3 years
Excited to interview Bill Inmon!
@databricks
Databricks
3 years
Nope – your eyes are not deceiving you! 👀 @InmonBill , the Father of the Data Warehouse, is speaking at #DataAISummit . Don't miss his fireside chat with Databricks CEO @alighodsi on the evolution of data architectures. Register now!
Tweet media one
0
12
36
1
8
32
@alighodsi
Ali Ghodsi
8 months
@laurenbalik Aww, this changes my Guy Fawkes Data Mascot plan for DAIS 2024. ☹️
0
1
32
@alighodsi
Ali Ghodsi
3 years
This is a big deal. Pandas, which Dara scientists love, just now works automagically on Spark on large datasets.
@matei_zaharia
Matei Zaharia
3 years
There's a lot of great functionality in #ApacheSpark 3.2, which just came out. #pandas API on Spark, RocksDB state store for streaming, and adaptive query execution are some of the highlights. Read our summary here:
1
23
100
1
2
32
@alighodsi
Ali Ghodsi
3 years
Join us on Boss Talk to discuss WFH, return to office, future of work...
@bhorowitz
benahorowitz.eth
3 years
I'm discussing “Boss Talk” with @alighodsi , @arielcoco , Felicia Horowitz, and a16z. Tuesday, Jun 8 at 5:00 PM MDT on @clubhouse . Join us! This week, we will talk about going back to the office and business travel with TripActions CEO Ariel Cohen
4
7
47
2
3
30
@alighodsi
Ali Ghodsi
1 year
Pretty cool! Now you can have LangChain agents use Spark!
@LangChainAI
LangChain
1 year
🧨 Spark SQL 🧨 One of the most powerful use cases we've seen is enabling agents to analyze data with SQL. Until recently you couldn't do this within Spark, one of the most powerful data engines. Thanks to @gengliangwang 's Spark SQL Agent, now you can!
Tweet media one
5
63
285
0
5
30
@alighodsi
Ali Ghodsi
3 years
Great podcast by @DavidGeorge83 on @20vcFund . Probably the most misunderstood thing about companies is sizing their TAM. "What's your TAM size?" -> people just takes a bunch of adjacent markets in IDC, add them up. Can be very misleading...
4
3
27
@alighodsi
Ali Ghodsi
2 years
@mim_djo See difference below. It's Parquet on a lake accessed with Photon (l). Data warehouses (r) accessing the Parquet files directly. The difference is 30x in price/perf with the multi-cloud edw on the right. Took us since 2017 to get this perf. Others can too, will take time.
Tweet media one
3
1
28
@alighodsi
Ali Ghodsi
3 years
Join @bhorowitz and me as we interview @davidmcj on Clubhouse now!
0
2
28
@alighodsi
Ali Ghodsi
3 years
@martin_casado (2/2) Why? Product founders are worried they'll become sales driven, culture will change etc. So they postpone the inevitable, which eventually leads to the board pushing for replacing the CEO with someone that gets enterprise GTM.
6
1
24
@alighodsi
Ali Ghodsi
2 years
It is a big deal!
@mim_djo
Mim
2 years
This is big, apparently you can now write Delta Table using just Python No Spark is required, this is absolutely a game changer. Great job Databricks!!!
4
2
51
1
8
23
@alighodsi
Ali Ghodsi
3 years
I had a lot of fun talking to @HarryStebbings at @twentyminutevc . Check out the podcast: #databricks #ai #data
2
2
23
@alighodsi
Ali Ghodsi
2 years
@mim_djo Don't worry... it's coming... we're moving fast.
3
2
19
@alighodsi
Ali Ghodsi
2 years
Super excited to grow this community.
@databricks
Databricks
2 years
. @DeltaLakeOSS just got even better. Meet #DeltaLake 2.0, now *entirely* open source. @michaelarmbrust shares how the latest features improve performance & manageability. #DataAISummit
Tweet media one
2
40
90
1
5
19
@alighodsi
Ali Ghodsi
9 months
@mim_djo We agree with your morning thoughts.
0
2
18
@alighodsi
Ali Ghodsi
3 years
@martin_casado Agree with both. Don't skimp on sales compensation? Yes sales folks are expensive, but it's for a reason.
3
1
19
@alighodsi
Ali Ghodsi
3 years
Join us in a few mins...
@bhorowitz
benahorowitz.eth
3 years
I'm discussing “Boss Talk” with @alighodsi , @pmarca , @FeliciaHorowitz and a16z. Tomorrow, Mar 16 at 5:00 PM PDT on @joinclubhouse . Join us!
0
4
38
2
0
17
@alighodsi
Ali Ghodsi
3 years
Interesting read about the early innings of the cloud...
@DanRose999
Dan Rose
3 years
I was at Amzn in 2000 when the internet bubble popped. Capital markets dried up & we were burning $1B/yr. Our biggest expense was datacenter -> expensive Sun servers. We spent a year ripping out Sun & replacing with HP/Linux, which formed the foundation for AWS. The backstory:
409
7K
29K
0
3
16
@alighodsi
Ali Ghodsi
3 years
Join in and listen in to our chat!
@protocol
Protocol
3 years
In 30 minutes! Join @JoePWilliams31 for The Inside View with @databricks CEO @alighodsi where he’ll get the inside scoop on what’s shaping up to be one of the biggest IPOs of the year. Join us here:
Tweet media one
0
1
5
0
3
17
@alighodsi
Ali Ghodsi
3 years
Join us now to listen to Boss Talk at Clubhouse with @bhorowitz @FeliciaHorowitz @pmarca
@bhorowitz
benahorowitz.eth
3 years
I'm discussing “Boss Talk” with @alighodsi , Felicia Horowitz, and a16z. Tomorrow, May 11 at 5:00 PM PDT on @clubhouse . Join us! please reply to this tweet with questions you’d like us to answer
4
8
46
2
1
16
@alighodsi
Ali Ghodsi
3 years
Join in! It’ll be fun.
@bhorowitz
benahorowitz.eth
3 years
I'm discussing “Boss Talk with Sanjit Biswas” with @alighodsi , Felicia Horowitz, @pmarca , Sanjit Biswas, and a16z. Today, Apr 27 at 5:00 PM PDT on @joinclubhouse . Join us!
5
13
62
1
2
16
@alighodsi
Ali Ghodsi
3 years
Join our conversation today!
@databricks
Databricks
3 years
Don't miss @AliGhodsi at @FirstMarkCap 's Data Driven NYC this afternoon with @MattTurck ! Follow along and submit your questions for us to discuss via #DataDrivenNYC . RSVP here:
0
6
15
0
3
15
@alighodsi
Ali Ghodsi
1 month
Hey @mattturck , it'll be easier to read the map if you start each box with: AWS, Azure, GCP!!
@mattturck
Matt Turck
1 month
It's out! After hundreds of hours of work, excited to publish the TENTH annual MAD (Machine Learning, AI & Data) Landscape. 🔥🔥🔥 The OG of data/AI market maps is back, bigger than ever lol + 24 themes we're thinking about in 2024 w/ @AmanKabeer11
23
82
370
1
0
14
@alighodsi
Ali Ghodsi
2 months
Super awesome @svangel !
@svangel
SVA
2 months
We all share responsibility for building AI that improves lives & unlocks a better future for humanity. @SVAngel makes this pledge, and we are proud to initiate this Open Letter: Build AI for a Better Future. Please join @OpenAI , @Meta @Google , @YCombinator , @HuggingFace ,…
10
30
103
0
1
11
@alighodsi
Ali Ghodsi
3 years
Join in, this will be fun!
@bhorowitz
benahorowitz.eth
3 years
Join us for “Boss Talk” as we talk about building a new media institution with Substack founders @cjgbest and @hamishmckenzie with hosts @alighodsi , @pmarca , and @FeliciaHorowitz Tomorrow, May 18 at 5:00 PM PDT on @clubhouse . Join us!
3
16
67
1
0
11
@alighodsi
Ali Ghodsi
3 years
Join us now and hear about M&A!
@divine4thletter
DIVINE
3 years
“Boss Talk M&A Edition” with @bhorowitz , @alighodsi , @eugenio_pace , @toddmckinnon , @FeliciaHorowitz , and @a16z . Today, Apr 20 at 8:00 PM EDT on @joinclubhouse !
0
1
6
1
0
9
@alighodsi
Ali Ghodsi
3 years
@martenmickos Thanks my friend. We are happy customers of H1, security is frankly top of mind more than ever for SAAS companies like us.
0
0
9
@alighodsi
Ali Ghodsi
3 years
@martin_casado Go To Market. Largely sales and marketing...
0
0
8
@alighodsi
Ali Ghodsi
1 year
Wonder how many of these could be soon cracked: In general, seems math would be revolutionized by approaches like this.
2
0
8
@alighodsi
Ali Ghodsi
3 years
@alex @matei_zaharia @CollisionHQ We gotta do it again then!
0
0
6
@alighodsi
Ali Ghodsi
1 year
Seems most theorem proving in the past was axiomatic. But LLMs understand natural language, which happens to be the language mathematicians use to convince each other in proofs.
2
0
6
@alighodsi
Ali Ghodsi
4 years
@vyas_sekar @justinesherry @notypes @fangyi_zhou_ I would find that while I was at the gym @vyas_sekar published a couple of more SIGCOMM papers.
0
0
6
@alighodsi
Ali Ghodsi
3 years
@_NelsonYap_ @SnowflakeDB DeWitt rule doesn’t let you name other vendors unfortunately. If you don’t have direct file access it will always be much slower.
0
1
4
@alighodsi
Ali Ghodsi
2 years
@martin_casado @jeanqasaur @sydgibs @frasergeorgew You learn a lot of useful things that help you as a CEO but it’s far from sufficient to ensure success too…
1
0
4
@alighodsi
Ali Ghodsi
1 month
@PrescientTrade Presumably a very small portion of the 200M paying users are coding.
1
0
4
@alighodsi
Ali Ghodsi
3 years
@kheimerl @klueska @amplab Lol, I cannot deny any of that.
0
0
2
@alighodsi
Ali Ghodsi
1 year
@peterabcz @TheOfficialACM Well deserved @peterabcz . Congratulations!
1
0
2
@alighodsi
Ali Ghodsi
1 month
@mattturck Not EVERY box!!!
1
0
2
@alighodsi
Ali Ghodsi
3 years
@martin_casado Big congrats @raghuraghuram , well deserved!
0
0
2
@alighodsi
Ali Ghodsi
2 years
@Triamus1 @DataMic It’s coming. Rest assured. They’ll be even better too, with great dashboarding experience from Redash.
1
0
2
@alighodsi
Ali Ghodsi
2 years
@mim_djo Actually @Scribd built the standalone library!
0
0
1
@alighodsi
Ali Ghodsi
1 year
@Grammarly is awesome!
@ClementDelangue
clem 🤗
1 year
People don't talk about it much but @Grammarly is one of my favorite applications of generative AI. Bigger than most companies out there both in usage & revenue, @huggingface customer for a few years now & with a very positive impact on the world! I use it everyday.
25
56
621
1
0
1
@alighodsi
Ali Ghodsi
4 years
@esammer Sorry for your loss Eric.
1
0
1
@alighodsi
Ali Ghodsi
3 years
@klueska @amplab Thanks Kevin! You got me permanently switching from skiing to snowboarding! :)
0
0
1