Dilip Kumar

building @polarisaixagent
tinkering with models meant for the edge
I use opencode btw

[Dilip Kumar @D_Leap07]

Gentlemen and Ladies, It is with great pleasure that I announce Glyph
A single platform for generating an illustration for your deep technical questions
Waitlist is now open : https://t.co/t7IWi1Uvng
Tried to generate a few algos this morning, here's KNN

[Dilip Kumar @D_Leap07]

Gonna start using Hono hereafter

[Dilip Kumar @D_Leap07]

@VazeKshitij blessed my morning!

[Dilip Kumar @D_Leap07]

Man i love figuring out errors on my own!
so liberating.

[Dilip Kumar @D_Leap07]

had no idea Smalltalk was an actual fkin programming language!

ffs its not automated man

[Dilip Kumar @D_Leap07]

@keerthanahq I ship software that I'd use everyday, not slop that works just once.

I'm aware this for a GTM role, just tryna grab your attention for AI Engineering, ML or Dev roles.

[Dilip Kumar @D_Leap07]

@merishabh_singh Just sent you a DM
Coming from Agents, Local Voice models, and building out of research papers from my last internship this is really exciting!

[Dilip Kumar @D_Leap07]

i feel sick man, absolutely no sleep.
it stings even more to know that i din make it to epoch 🥀

[Dilip Kumar @D_Leap07]

Gonna gring on Math Recap for ML today. I am the literal opposite to what i used to be back when i started.

[Dilip Kumar @D_Leap07]

@VazeKshitij lmk if you're free to meet sir! hella excited!

[Dilip Kumar @D_Leap07]

step 1 => build someting
step 2 => do some research on how to sell
step 3 => while doing so, realise HF has already built something similar

step 4 => give up

[Dilip Kumar @D_Leap07]

The goal isnt to write the best code there is, any agent can do that already
its to provide the convenience to run the models, test them, compare the numbers all in one platform.

checkout https://t.co/6s43iz9V2X right away!

[Dilip Kumar @D_Leap07]

Hello all
I built Polaris AI a platform to ingest any ML research paper to implement it end to end
The BIG BOLD DREAM is to nullify falsified numbers claimed by papers by implementing papers on a sandbox
Have a go youself at https://t.co/9sv906sFNv
anxious to hear your thoughts🧵

[Dilip Kumar @D_Leap07]

Github repo https://t.co/YI4h575F0i & in the future a single install npm registry to just install the code for the paper you need.

[Dilip Kumar @D_Leap07]

The larger goal is to have a collection of papers that have been successfully implemented. All models have benchmarks to vouch for its performance, the aim is to have the same tests for papers
Polaris is the huggingface for papers, Battle tested and implemented for anyone to use

[Dilip Kumar @D_Leap07]

This is motivated after being lost over reading papers going absolutely nowhere as a beginner there was no tool to guide me through the jargon
For Researchers there isnt a unified platform to both write code & run them on GPUs with a sandbox
until now Polaris is your north star🌟

[Dilip Kumar @D_Leap07]

@higgs_Divyansh @cneuralnetwork You don't even have to worry about that anymore, have a go at https://t.co/9sv906sFNv

[Dilip Kumar @D_Leap07]

Bringing the agent swarm to your browser, ML researchers deserve the perks of vibe coding too.
@PolarisAIxAgent

[Dilip Kumar @D_Leap07]

Bringing the agent swarm to your browser, ML researchers deserve the perks of vibe coding too.
@PolarisAIxAgent

[Dilip Kumar @D_Leap07]

While debugging figured out that the coding agent in Polaris' CODE agent would end up hanging in between.
No traces, Nothing on redis to show client.
Turns out the max-iter limit was being hit because the goal was too large. so the agent would retry over and over and exhaust.

[Dilip Kumar @D_Leap07]

I couldve just pushed a landing page with a waiting list and called it a day. Instead i've added couple more features and im debugging it with caffeine.
UPDATE:
Subscription tiers are added, i've been dying to add this feature from the other day, its coming up soon(GPU access)

[Dilip Kumar @D_Leap07]

UPDATE:
the agents are working, its really annoying to setup the github org authentication :/s