proxis

Proxis (YC S24) puts your enterprise email on autopilot. We build automated drafting and email sending for enterprise teams handling outbound.

[Sagar @sagredd]

Your inbox is out of control.

@proxis_ai auto-drafts replies to support tickets and sales pings by:

- Connecting to your Slack/CRM for context
- Matching your writing style
- Auto-sending high-confidence replies
- Drafting tricky ones for review

[proxis @proxis_ai]

10x Engineers at Proxis work on autopilot to keep your mailbox on autopilot.

Try: https://t.co/Vy16nvX6ZE

[Adi @pupscub]

Tonight’s the night. We’re so back

[proxis @proxis_ai]

Put your mailbox in auto-pilot, backed by @ycombinator
we generate context aware responses that sounds like you!

[Adi @pupscub]

@proxis_ai does it cheaper AND better than anyone

[Liam Collins @liamcollins____]

🏆 Congrats to @Sterling_Hooten who just won $100 🏆
The Proxis Website ROAST is a wrap!

[proxis @proxis_ai]

Some call it QA. We call it ROASTING.

Sharpen your keystrokes and brighten someone's linkedin doom scroll today.

We’re offering $100 to the funniest + most useful critique of our new website: https://t.co/xcGBYf9GAA. Winner announced on Friday June 6.

[Liam Collins @liamcollins____]

$100 Website Roast Challenge 🚨

We just launched our new website — https://t.co/UywGaaLuWZ — and we want your honest, hilarious, and brutally useful feedback.

💸 Prize: $100 to the funniest and most helpful roast.
What we’re looking for:

[Liam Collins @liamcollins____]

Wherever we go, there will be piñatas 🪅

For those in the Bay come join us tonight at the @proxis_ai ugly sweater holiday party🎅

Hosting with our friends from @jpmorgan, @googlecloud , and @wilsonsonsini

DM me for details.

[Liam Collins @liamcollins____]

NEW RELEASE

What does Llama 3.3 mean for you?

A few hours ago Meta released the latest Llama 3.3 model. Here are a few things you need to know:

- It is now available in the 70b size. This makes it feasible to run locally
- It's performance is comparable to 3.1 405b (see benchmark categories below) with more cost effective inference (~1/10th of the cost)
- It outperforms 3.1 405b on math and reasoning

Open source is here. If you're looking to fine tune and distill these models, reach out:
https://t.co/7nyW7v2pCC

[Liam Collins @liamcollins____]

The bet is on open source.

Last month I spoke at the Wharton Tech Conference about @proxis_ai and the future of open source models.

There is a transformational shift of enterprises transitioning toward open source models. There are a number of very good reasons for this:
- Customization and control (think fine tuning and distillation)
- OWNING your models!
- Privacy and Security
- No vendor lock in
...
___________________________________________________
Proxis (YC S24) is helping drive that transformation for enterprises and developers. Proxis is the the Vercel of fine tuning & distillation - the easiest way to spin up a custom, production-ready LLM in under 10 minutes. Hosting, privacy and security, and ownership are all sorted with the proxis platform.

If you are an enterprise AI leader or developer feel free to schedule a call: https://t.co/PYYeFD33tt
_____________________________________________________

Thank you to @Wharton , @whartontechclub for coming out to SF and for bringing together such a great group of builders.

@jacksonstokes_
@leonz4264

[Liam Collins @liamcollins____]

2 days to go!

Looking forward to seeing you all there.

We'll be launching our model optimization tool. We'll walk through the tool and discuss fine tuning and distillation techniques.

Video call link: https://t.co/PlFkswvNuM

[proxis @proxis_ai]

Ok so we really don’t want that, how do we correct this? Models can be aligned to human values using reinforcement learning, supervised fine-tuning and human feedback. I’ll dive into these in greater detail in the next post.

[proxis @proxis_ai]

👁️What is AI Alignment? 👁️

It has been a hot topic recently, but it’s not immediately obvious what it means.

Simply put, AI alignment means models act in ways humans intend them to. It ensures AI models are aligned with human values, intentions, and ethics.

More in 🧵

[proxis @proxis_ai]

What does it look like when models are misaligned with our values? We might see model outcomes that:

- Unintentionally leak or misuse sensitive information

- Spread misinformation

- Act on unsafe autonomous decisions

- Recommend illicit activities

- you get the point...

[Liam Collins @liamcollins____]

If you haven't yet heard of these companies you soon will. Here are some YC founders to keep an eye on.

@covaldev

@wordware_ai

@FINNYAI

@dataleapHQ

@plume_energy

And of course @proxis_ai

[proxis @proxis_ai]

Proxis presenting the new VSCode kernel optimizer at the YC Launch.

Check out the launch as well as other model optimization tools here:
https://t.co/xcGBYf9GAA

Join the conversation on discord and get off the waitlist sooner:
https://t.co/iwHA4Le1xo

[proxis @proxis_ai]

Kernel level optimizations? We got you. Coming soon at: https://t.co/3aQqEePk94

Big thanks to the OpenAI team.

[Y Combinator @ycombinator]

Congrats to the winners of the @OpenAI x YC o1 Hackathon!

@liam_a_collins & @jackson_stokes from @proxis_ai
@_maximall & @TaiebBennani from @vera_health_ai
@AryaBastani & @KeatonElvins from @camfer_inc

[Y Combinator @ycombinator]

YC S24's @proxis_ai is the first dedicated platform for LLM distillation and serving, unlocking production-ready models at 1/10th the cost.

Congrats on the launch, @liam_a_collins and @jackson_stokes!

https://t.co/NITWdN2wpa