Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud

Enterprise & production-focused opensource LLM SDK.
Scale on the edge with a lang (LLM) chain serverless cloud.
Compliance friendly.
#Reactive #Java #Graal

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

Love your LLM!
#LLMPride

[sankalp @dejavucoder]

good prompt engineers thank their LLMs

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

@ashugarg @sama @eshear @FoundationCap @JayaGup10 @VinIyengar https://t.co/8q81KFLQYs

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

hey #KloudKarnival we are here for CNCF. Say hi to us - we do a LOT of work on Kubernetes, Jsonnet, Lambda, AWS....and LLM ChatGPT chains.

Made with love in Delhi.
https://t.co/8q81KFLj8U

[CNCF New Delhi @cncfnd]

🎉 Tomorrow is the big day! Join us for an unforgettable journey at #KloudKarnival. 🚀 Here's the agenda:

📅 Date: Sunday, 03rd Sept 2023
🕒 Time: 01:30PM
📍 Venue: Tata1mg Gurgaon

Note : Beat the rush! Please arrive a bit early for smooth check-in at #KloudKarnival #cncfnd🎉

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

@whoisnnamdi happy to help any of your companies with GenAI devtools.
We think that EdgeChains generally 10xes go-live on GenAI (as compared to the other chains frameworks).

This is why you should think of it - https://t.co/LCS3tADewT

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

@thibauld @X @Support @StephNass @OpenVC_ support @stephnass

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

@arakooai will be at @SnowflakeDB's #DataCloudWorldTour tomm.
Will be showcasing open source EdgeChains and how you can build your own #GenerativeAI webapps. Interoperates seamlessly with Snowflake.

First 5 people that talk to me tomm get free dev time as well.
See u tomm!

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

@siddharth_hacks @hackthisfall happy to help. we are an open source org that builds generative AI sdk in India.

For beginner developers in AI/chatGPT/LLM.

[Arakoo Devs @arakoodev]

So few AI hackathons happen online. @soonami_io is doing a great thing with this hackathon.
Upto 500K as investment if u make it to the top.
And here's a second offer - if u want technical help with ur AI app, @arakooai will help - Edgechains is fantastic to take AI to production

[Arakoo Devs @arakoodev]

@solomonstre Prompt-Engineering as Code.

Edgechains is built on top of jsonnet and the entire chain is modeled as a configuration on top of it.

node = LLM
container = chain
service = prompt

https://t.co/QgmLria9RM

[Arakoo Devs @arakoodev]

This is so true and that is why we built @arakooai on top of jsonnet.
Our belief is that generative AI applications = prompt X chain X LLM config management problem.

We think the right way to model this problem is exactly how Helm/Terraform model containers X nodes X services

[swyx @swyx]

just as i was tweeting, OpenAI now elegantly supports the LLM Core / Code Shell pattern: https://t.co/ZpSmXpVRZM

OpenAI is very much pushing the "God Model centric" view of the world (altho @realGeorgeHotz had great commentary on what is rumored to be under the hood of GPT4 in our podcast). Prompt engineering is a kludge, one model can handle all languages, all modalities, all preferences.

I wonder what the inverse LLM Shell/Code Core world would look like in a world of ubiquitous @smolmodels.

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

LLM meets Kubernetes/Borg config management meets serverless = EdgeChains.

[ethan ding 📊 @TheEthanDing]

Unqualified opinion weakly held incoming, but I don't think LangChain and LlamaIndex are set up for success - it reminds of this fantastic article: https://t.co/kieBkvMrjx

how can we know what LLM infrastructure is going to look like when no one has built a good LLM application outside of raw OpenAI?

Facebook and Uber don't use Snowflake - they use their own warehouse internally, and products like Snowflake exist because the need for Snowflake was validated by applications that first validated the need for big data.

Who's using these tools today to build big things? We use langchain-like architecture internally, but everything at @textql is done in-house. I'd be shocked if companies like @_hex_tech or @gleanwork use LangChain under the hood.

My prediction:
1. someone's going to build an Uber-sized app on top of LLMs with custom infra
2. someone on the platform eng team is going to start a company that replicates that custom infra
3. that company will outcompete Langchain commercially within 2 years

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

OpenAI hackathon this weekend!

If you want to build production-stable & scalable chains in this hackathon, use EdgeChains. When you use edgechains, everything you do is automatically exposed as an API.

Remember VCs invest in products that can scale.

https://t.co/8q81KFLj8U

[Stellaris Venture Partners @Stellaris_VP]

⌛ Last date to register for the GenAI Hackathon: AI for India is tomorrow, 6th June.

Form your teams of 1-4 & apply today for an opportunity to solve real-world problems with AI, under the expert mentorship of Stellaris & @join_ef

Apply now: https://t.co/hTsgCiiqTm

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

@TIIuae anyone from the Falcon-40B team available next week to meet ? we are traveling from India to showcase EdgeChains (https://t.co/8q81KFLj8U) in Dubai.

Would love to meet.

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

this is very very true. how many banks now have chatbots powered by GPT ?
Cos production is still hard. Building devtools with a production focus is a very different set of tradeoffs than building for data scientists in jupyter.

that's what we are hard at work @arakooai

[Timothy Chen @tnachen]

But building a production product with LLM is still very hard, as you run into many issues that’s described here (https://t.co/nClSTJ2MYJ)

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

Second one is in Bangalore with Stellaris.

Same offer applies to both - we will assist you on everything from algorithm to implementation to servers, if ur using the EdgeChains SDK.

https://t.co/s8mGTsFAyP

[Stellaris Venture Partners @Stellaris_VP]

Inviting exceptional developers, data scientists, and AI enthusiasts to join our first GenAI Hackathon💫alongside @join_ef.

Hone your AI skills, collaborate with like-minded people, and start using your technical skills to make an impact👨‍💻👩‍💻

Register: https://t.co/hTsgCiiqTm

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

If you want to participate in Generative AI hackathons in India, there are TWO happening this weekend.

Our offer - We will help you with GPU credits, servers, etc (as long as you use the EdgeChains SDK https://t.co/8q81KFLj8U)

First one is in Mumbai

[Maharashtra State Innovation Society @MSInSociety]

Mumbai's Hackathon is happening on June 3&4 at WeWork, Andheri. Build for the future on the Generative AI track, make our digital world safer on the Cybersecurity track, or make our beloved city a better place to live in on the Solve for Mumbai track
Link:https://t.co/qTPa1QCofV

[Jaya Gupta @JayaGup10]

More quotes from @joannezchen :

“Having a moat means having a customer for the next 3 years.”

“Open source is great for two reasons, you get data around consumer behavior and for the free engineering resources”

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

As context windows grow to 100k and beyond...the need for vector DB drop. Instead, the focus shifts to techniques like "FrugalGPT" for chain-of-thought and prompt compression.

FrugalGPT will be in the next release of EdgeChains.

[Sandeep Srinivasa @sandeepssrin]

ur right, but u dont need to use vector db to reduce cost. you can use FrugalGPT to do that.
so the larger context window does eliminate the need for vector DB.
we will still need tools like gpt_index, edgechains/langchain for composing the chain-of-thought.

https://t.co/r6omA64OOb

[Arakoo.ai : #EdgeChains & Lang (LLM) Chain Cloud @arakooai]

@mckaywrigley @Replit @langchain would love to show you what we are doing with edgechains - deployment & scale friendly chain dev SDK!
https://t.co/8q81KFLj8U