Eshan

@esh

22 Following

89 Followers

Eshan pfp

I'm excited to share our whitepaper on decentralized, verifiable AI training! This paper outlines our fully scalable training with real time verification. At Impulse AI, we believe that 100% verifiability is required to make decentralized training attack proof 🔒🧠🚀 https://github.com/impulse-ai/whitepaper/blob/main/Impulse_AI_Whitepaper.pdf

0 reply

0 recast

0 reaction

Eshan pfp

Fully verifiable distributed training announcement coming tomorrow 🤯🚀

0 reply

1 recast

0 reaction

Eshan pfp

EthCC takeaways - stablecoins and more stablecoins. heard lots of cool mechanisms - never heard so many people talk about revenue before in crypto. loved it - Cannes is a great city to host a conference. everything is walkable and i ran into people on the street - lot less "AI x crypto" events than ETH Denver and other conferences. probably cuz we need to build product now - defi is back in a different and interesting way - lot of cool founders in town

0 reply

0 recast

0 reaction

Eshan pfp

How we use AI at Impulse 🤔 When I started Impulse AI, we needed to build a functional frontend for the product. We didn’t have capacity on the team to build the frontend ourselves, so I started reaching out to contractors on Upwork. The estimates were somewhere between $3,000 to $5,000. The plan was to get the frontend built, and then our engineering team would take over to integrate it with our backend systems. Derek Anderson (thank you so much 😀) introduced me to v0 by Vercel. I figured, why not give it a shot? My process was fairly simple: For each page, I'd prompt for the basic layout, then focus on reducing unnecessary functionality to keep it lean, fix the user experience, and finally work on styling. What I built wasn't the most visually stunning application out there, but it’s intuitive, easy to use, and readable.

1 reply

0 recast

0 reaction

Eshan pfp

💡 We've just launched our LLama 3.1 8B tutorial in our docs! The tutorial covers everything you need to know about: - Installing and using the Impulse AI SDK - Preparing your data for optimal fine-tuning - Launching your first fine-tuning job 👉 https://docs.impulselabs.ai/introduction

0 reply

1 recast

0 reaction

Eshan pfp

Llama 3.1 8B is available for fine-tuning! DM me if you want to try the product out. I’m happy to help you get started! Try our app at app.impulselabs.ai Docs: docs.impulselabs.ai

0 reply

0 recast

0 reaction

Eshan pfp

Impulse AI now fully supports fine-tuning Llama 3.1 8B! 🦙 - Both base & instruct models supported ✅ - Simplified fine-tuning, focused on your data ✅ - Get started fast with our SDK & tutorial ✅ Read More 👇 https://www.impulselabs.ai/blog/unlock-the-power-of-llama-3-1-8b-with-impulse-ai Try https://app.impulselabs.ai

0 reply

0 recast

0 reaction

Eshan pfp

⭐️⭐️⭐️ I’m excited to share that I’ve started Impulse AI! Impulse is an AI training and inference platform that makes it easy to fine-tune and deploy open source LLMs. We’ve been hard at work the last month building out our SDK, web app, and AI platform! Stay tuned over the next few weeks for announcements on new features, our technical whitepaper, and partnerships. AND we’re offering free credits for the first 100 users who sign up on the platform 🙏 Learn more at http://impulselabs.ai 🚀

0 reply

0 recast

0 reaction

Eshan pfp

hello 🌎 Impulse AI is a decentralized AI training, fine-tuning, and inference platform! check us out at https://www.impulselabs.ai/

0 reply

0 recast

0 reaction

Eshan pfp

Something about decentralization that is rarely mentioned: As more users and contributors contribute to a product, it's more likely we end up building a better product, resulting in more value creation

0 reply

0 recast

1 reaction

Eshan pfp

I'm like a month too late on this, but still a fun read https://vitalik.eth.limo/general/2025/01/05/dacc2.html

0 reply

0 recast

0 reaction

Eshan pfp

turns out tarrifs + subsidies does help american manufacturing https://www.noahpinion.blog/p/yes-reshoring-american-industry-is

0 reply

0 recast

0 reaction

Eshan pfp

worth a read https://www.noahpinion.blog/p/the-players-on-the-eve-of-destruction

0 reply

0 recast

0 reaction

Eshan pfp

6 best schemers, play designers, and play callers in football right now (in no order) - andy reid - kyle shanahan - sean mcvay - ben johnson - matt lafleur - kevin o'connell each offense is so unique

0 reply

0 recast

0 reaction

Eshan pfp

Anyone who compares and contrasts Japan vs America needs to read this https://www.noahpinion.blog/p/yes-americans-are-much-richer-than

0 reply

0 recast

0 reaction

Eshan pfp

if gusto had a PEO, I would not move away. trying out other PEOs software has given me a new found appreciation for how good the UX on gusto is

0 reply

0 recast

0 reaction

Eshan pfp

the ringer's new website is terrible and so hard to use. it's like they took all the best UX practices and did the exact opposite to "make it look cool"

0 reply

0 recast

0 reaction

Eshan pfp

🔥 Important update in the world of decentralized AI Below is a demo of our decentralized ML training protocol working! Data centers and others will be able to supply compute to the network, and run training jobs for a fee based on the requirements of the training / fine-tuning job. This is the first step to launching a fully decentralized, open source, permissionless network, but it's supremely important. Our small team of 4 has now shipped our LLM Fine-Tuning SDK and our devnet within a month of each other. Who says small teams can't accomplish big things 😃 Carpe Diem!

0 reply

0 recast

0 reaction

Eshan pfp

We’ve launched the devnet of our decentralized machine learning training protocol! This is the first step in our vision to help decentralize AI and make it easy, safe, and affordable for everyone to build and use AI. To learn more , please visit: https://www.luminolabs.ai/blog/lumino-launches-decentralized-ml-training-protocol

0 reply

0 recast

0 reaction

Eshan pfp

We’ve launched Lumino on Product Hunt! You can fine tune models without having to deal with infra setup for training. We’re also up to 80% cheaper than AWS/GCP. And best of all, you can fine-tune directly from our UI, no coding required! Check it out: https://www.producthunt.com/posts/lumino-ai

0 reply

1 recast

2 reactions