Varun Vummadi Profile
Varun Vummadi

@varunvummadi

868
Followers
191
Following
12
Media
161
Statuses

CoFounder and CEO @GigaML (YC S23)

San Francisco, CA
Joined March 2021
Don't wanna be here? Send us removal request.
@varunvummadi
Varun Vummadi
9 days
Every one at @ZeptoNow team is so excited to work with Giga ML they are sitting with us on the weekend for 8 hrs to scale our solution. Excited to work with the fastest growing company in the world.
Tweet media one
16
19
636
@varunvummadi
Varun Vummadi
5 months
Garry is the GOAT.He helped us get an O1 visa with recommendation letter and he is now allowing us to do all the events this batch because we missed last batch due to visa issues.He deeply cares about people and wants us to be successful
@alexrkonrad
Alex Konrad
5 months
"YC feels very much back." - OpenAI CEO @sama "A real champion of founders, and a true believer." - Airbnb CEO @bchesky "Pause the politics and focus." - an anon YC founder Inside Garry Tan's first year atop the world's best-known startup accelerator 👀
5
20
117
3
7
101
@varunvummadi
Varun Vummadi
9 days
10:20 PM and still going thanks @iamnishitraj for coordinating. We are going full scale today
Tweet media one
2
2
44
@varunvummadi
Varun Vummadi
6 months
Waiting to get groq cloud access just applied for @GroqInc crushes every other provider in throughput
@sundeep
sunny madra
6 months
Zoom out a bit 😉
Tweet media one
5
9
64
0
2
9
@varunvummadi
Varun Vummadi
1 year
@Culture_Crit AR/VR + AI will replace colleges in future
0
1
8
@varunvummadi
Varun Vummadi
15 days
This is my first time visiting Singapore one of the most cleanliest cities I have ever seen.
0
0
7
@varunvummadi
Varun Vummadi
13 days
If he really does that at least a million of Indians would be grateful to him through their life. The amount of struggle we faced to get in to US even after getting into YC our B1/B2 got rejected twice and finally the only visa option for is O1 and super grateful that we got
@theallinpod
The All-In Podcast
1 month
. @realDonaldTrump on granting automatic Green Cards to college graduates: "... What I want to do and what I will do is, you graduate from a college, I think you should get automatically, as part of your diploma, a Green Card to be able to stay in this country." "And that
555
810
6K
1
1
7
@varunvummadi
Varun Vummadi
2 months
One of the best google Product right now is @Waymo but sadly very few people get to experience it
1
0
6
@varunvummadi
Varun Vummadi
9 days
@_vrshny @ZeptoNow To be clear Zepto is the fastest growing company. We are not we are excited to work with them
1
0
6
@varunvummadi
Varun Vummadi
2 months
One of the best low calorie (10) high caffeine energy drinks I've ever had in my life. You can literally code for 24 hours straight after drinking these.
0
0
6
@varunvummadi
Varun Vummadi
12 days
Is there any time in history @airindia has departed at the time it is actually supposed to depart. Please reply if you had a time flight with air India
3
0
5
@varunvummadi
Varun Vummadi
8 months
Finally got my O1 visa today.
0
0
5
@varunvummadi
Varun Vummadi
9 months
@garrytan This is a brute force solution you can run mistral 7b on Mac Book using llama.cpp and run sentence transformers embedding model for rag locally and connect it with your email boxes using API like Gmail API.
1
0
5
@varunvummadi
Varun Vummadi
9 months
@sama Thanks Sama for literally changing the world starting the AI revolution with such great products
0
0
4
@varunvummadi
Varun Vummadi
9 months
Today @OpenAI gave hope to all the startups.
1
0
5
@varunvummadi
Varun Vummadi
4 months
Any one attending Nvidia GTC just making sure to not to miss great folks please reply if you are attending So that many folks like me would know that you are attending. #NVIDIAGTC
1
0
4
@varunvummadi
Varun Vummadi
9 months
This makes total sense considering the pricing and latency.
@Teknium1
Teknium (e/λ)
9 months
FYI the new code diffusion model paper by some people at Microsoft claims ChatGPT-3.5-turbo is 20B params..
Tweet media one
31
126
785
0
0
5
@varunvummadi
Varun Vummadi
1 year
Search feature for Chat GPT plugins would be super helpful for new plugins to be discovered when we can expect it @OpenAI @gdb @sama @tszzl
2
1
4
@varunvummadi
Varun Vummadi
5 months
If you are using vllm simply switch to tensor rt by @nvidia to increase the speed by 40%-50%
0
0
4
@varunvummadi
Varun Vummadi
7 months
Thanks for covering us @Kyle_L_Wiggers
@Kyle_L_Wiggers
Kyle Wiggers
7 months
Giga ML wants to help companies deploy LLMs offline
0
0
1
0
0
4
@varunvummadi
Varun Vummadi
1 year
@aleofriend @garrytan Yes we provide on premise deployment for your specific proprietary data think of it as making an LLM similar to Bloomberg GPT for your firm reach me out at varun @giga .do for more info.
1
1
4
@varunvummadi
Varun Vummadi
15 days
@jaltma Cold email is rarely working now a days after people started using AI to write cold emails the amount of email volume has gone up a lot. Got a 1.3 percent rejection emails in my first 3000 mails and 0 percent who accepted the offer or replied me with would like to learn more.
0
0
4
@varunvummadi
Varun Vummadi
1 year
Mail to founders @giga .do for our APl access. Our fine-tuned models outperform GPT4 for your specific use (like LLM for finance, insurance, health care, and enterprise data) and Support on-premise deployment for data privacy.
@garrytan
Garry Tan
1 year
Do you want to outperform GPT-4 by training your own models built off of open source LLMs? This seems like a no-brainer. Who wants to get locked in? Giga ML (YC S23) is tackling a *very* important problem.
Tweet media one
11
10
140
0
0
3
@varunvummadi
Varun Vummadi
5 months
@Teknium1 I can make an intro
0
0
3
@varunvummadi
Varun Vummadi
1 year
I think we are coming close to the day we can talk with Aristotle and steve jobs.
@timhwang
Tim Hwang
1 year
steve jobs, 1983
Tweet media one
9
50
245
0
0
3
@varunvummadi
Varun Vummadi
6 months
@Vishruth_N @amit_sethi I agree with this but I do think companies like Anthropic or Mixtral or similar can compete with hundreds of millions of dollars of funding but the issue is companies like Krutim with lot of funds and no engineering talent just launch a shit product with hype as nationalism it
1
0
3
@varunvummadi
Varun Vummadi
9 months
Thank you this is a massive help for the open source community.
@togethercompute
Together AI
9 months
We are excited to release RedPajama-Data-v2: 30 trillion filtered & de-duplicated tokens from 84 CommonCrawl dumps, 25x larger than our first dataset. It exposes a diverse range of quality annotations so you can slice & weight the data for LLM training.
Tweet media one
20
284
1K
1
0
3
@varunvummadi
Varun Vummadi
2 years
@Strandjunker @elonmusk Then How is the business model supposed to fuc****ng run and he is mentioning that ads are not enough for them if you can't extract at least $8 value from twitter with your followers It makes sense for you to leave.
0
0
3
@varunvummadi
Varun Vummadi
1 year
@ylecun This is awesome
0
0
3
@varunvummadi
Varun Vummadi
6 months
Is azure down just for us or for every one else @AzureSupport @Microsoft
2
0
3
@varunvummadi
Varun Vummadi
1 year
@garrytan Thanks a lot Garry for helping us reach more people
0
0
3
@varunvummadi
Varun Vummadi
1 year
@RayDalio I think it is extremely hard to get ultra valuable criticism even from close friends .Most of people don't care about us or don't want to hurt us. The 1st kind will give the honest valuable criticism rather than the 2nd type .Any tips on getting valuable criticism?
0
0
2
@varunvummadi
Varun Vummadi
7 months
@DrJimFan There are only popular models from big companies in the chat bot arena I think it should be like hf leader board and should test top models in arena elo.
0
0
2
@varunvummadi
Varun Vummadi
1 year
@sdianahu @calvinfo @dessaigne @RichAberman @koomen I want to thank you for helping founders like us with such great content really thanks a lot.
0
0
2
@varunvummadi
Varun Vummadi
1 year
@ycombinator Daltons and Michael yt series is awesome thanks for helping us with great advice
0
0
2
@varunvummadi
Varun Vummadi
1 month
@greg16676935420 I was super surprised to know this when I first opened the chase account when I moved to US. The reason might be physical branches I guess people are comfortable keeping money in banks where they can actually visit the bank
0
0
1
@varunvummadi
Varun Vummadi
1 year
@HusseinYahfoufi @OpenAI @gdb @sama @tszzl That's impressive but it would be great if there is a link to click on it to redirect us to that plugin.
1
0
2
@varunvummadi
Varun Vummadi
1 year
Adios IIT Kharagpur
Tweet media one
0
0
2
@varunvummadi
Varun Vummadi
2 months
We can see 3d objects in the map earlier it just used to be 2d
0
0
2
@varunvummadi
Varun Vummadi
1 year
@Kaden_Lane_SF @garrytan We are using RAG + fine tuning for best performance our fine tuned models will be updated every 3 - 6 months to stay on TOP since our models have low parameters fine tuning is comparatively cheap
0
0
2
@varunvummadi
Varun Vummadi
1 year
@tunguz True 🤣🤣
0
0
2
@varunvummadi
Varun Vummadi
9 months
@ghatageanup @garrytan That's true I totally forgot 😅 about the vector store pg vector is solid vector db.
0
0
1
@varunvummadi
Varun Vummadi
3 months
@deepit_patil @kabirnagral I just realised that you both are using orange back grounds
1
0
2
@varunvummadi
Varun Vummadi
1 year
One of the best guides to learn prompting on GPT4
0
0
1
@varunvummadi
Varun Vummadi
5 months
@raghavar0 @KrutrimAI They might have fine tuned it on open AI conversation data which they scraped from hugging face I guess. ( Which they are not allowed to legally if they are building competing models with open Ai)
0
1
2
@varunvummadi
Varun Vummadi
1 year
@ashah0052 @cornell_tech We don't need fact gpt for Wikipedia data we need it for logical questions that is where GPT models go terribly wrong.The chance that GPT gets an Wikipedia question wrong is super low .
0
0
2
@varunvummadi
Varun Vummadi
1 year
@rpoo How neural networks exactly work rather than abstraction. If you know you should be able to write a math function which gives a similar output
0
0
2
@varunvummadi
Varun Vummadi
2 years
This is an awesome resource for any one who wants to learn about GPT after Chat GPT and understanding the GPT model Thanks to @karpathy for posting awesome content
0
1
1
@varunvummadi
Varun Vummadi
1 year
@gdb How can it add up over time can you elaborate just shortening an URL how is it going to help ?
0
0
1
@varunvummadi
Varun Vummadi
2 months
@leojr94_ @surfacelabs_ai and @saharsh_wal might be able to help you with lead qualification
1
0
1
@varunvummadi
Varun Vummadi
1 year
@ThePeterMick 0 percent
0
0
1
@varunvummadi
Varun Vummadi
9 months
@deliprao Have you tried their model it is faster and better than GPT4 for code specific use cases it is easy to shit on some one rather than actually building stuff. How can you evaluate LLM on code rather than human eval?.
1
0
0
@varunvummadi
Varun Vummadi
1 month
@MohapatraHemant @airindia @SingaporeAir We ( Me and my cofounder) also had a bad experience with @airindia couple of months ago will never buy a ticket from them again. We have never tried @SingaporeAir but we love @cathaypacific the layover in hong kong splits flying time in three best way.
0
0
1
@varunvummadi
Varun Vummadi
1 year
@remotejoeclark Azure is the best platform for at least AI startups their Open AI service is the best in the market.
1
0
1
@varunvummadi
Varun Vummadi
1 year
There is a high chance that this happens to everybody who is doing low level jobs if you don't somehow enter into the AI industry there is a super high chance that your jobs will get replaced and you can't compete with people who have high computing power in any way.
@sama
Sam Altman
1 year
a new version of moore’s law that could start soon: the amount of intelligence in the universe doubles every 18 months
1K
2K
15K
0
1
1
@varunvummadi
Varun Vummadi
1 month
@SanchetiBindu You should try flying to Dubai at night Dubai is one of the cities with best views from plane
1
0
1
@varunvummadi
Varun Vummadi
1 year
@pmddomingos Yes we need it to write emails a lot of times as a student I need to write boring mails to professors for research and acceptance and even for applying jobs customised to them chat gpt makes it super simple
0
0
1
@varunvummadi
Varun Vummadi
1 year
@daltonc @ycombinator When we are trying to explain our startup in less than 50 chars in the For YC application we struggled a lot and changed it at least 30 times and were not satisfied at the end .The pitch is "Giga is an AI driven personalised learning platform". How to improve this skill ?
0
0
1
@varunvummadi
Varun Vummadi
1 year
If you are a student who is looking for front end developer or UI/UX designer interns make sure to check out this opportunity. #opportunity #frontenddeveloper #ux #ui #designer
0
0
1
@varunvummadi
Varun Vummadi
6 months
@Vishruth_N @amit_sethi This is actually happening in the space right now every one is literally competing for this companies like together Ai keeps reducing the prices
1
0
1
@varunvummadi
Varun Vummadi
1 year
Yep GPT4 got worse
@matei_zaharia
Matei Zaharia
1 year
Lots of people are wondering whether #GPT4 and #ChatGPT 's performance has been changing over time, so Lingjiao Chen, @james_y_zou and I measured it. We found big changes including some large decreases in some problem-solving tasks:
Tweet media one
123
796
3K
1
1
1
@varunvummadi
Varun Vummadi
2 months
@peer_rich I thought this was an ad for for a second.
0
0
1
@varunvummadi
Varun Vummadi
1 year
This video keeps popping up on YouTube even if I click not interested or do not recommend channel it is like YouTubes Algorithm wants to support these scams. @YouTube
Tweet media one
1
0
1
@varunvummadi
Varun Vummadi
2 months
@buccocapital Giga ML is the knife
0
0
1