Batuhan Ayci

Postdoctoral Fellow @MSHSNeurology | MD @cerrahpasatipf | AI geek | 2 x GB10 | Views my own

[Batuhan Ayci @batuhanayci]

A 3D neuroanatomy atlas that runs in a browser. Everything is in MNI152 space, the same coordinates as the template MRI, so the model and the scan correspond and you can move between them.

https://t.co/4bCy6eMjAW

[Özgür Kara @ozgurkara99]

1/8 BudgetPix: one model, any compute budget.

Our pixel-space image generator adapts computation to image content at inference time. Same prompt, same noise, from 100% down to just 10% of the tokens.

→ Project Webpage: https://t.co/aDxG1ZDq7V

[Batuhan Ayci @batuhanayci]

@MiaAI_lab @TechMDAI @NVIDIAAI Solid work, I am not 100% sure if it is worth to occupy two sparks if I can get similar speeds with less cache with just one spark

[Batuhan Ayci @batuhanayci]

Not something you see every day. Today we're saying goodbye to our 7T MRI 🧲🏗️

[Batuhan Ayci @batuhanayci]

@klctutku Yanar

[Batuhan Ayci @batuhanayci]

I feel lucky…

[Batuhan Ayci @batuhanayci]

Time to get out of the shadows. 🕶️

Meet my compute: DGX Spark + ASUS GX10. 256GB of unified memory, fully local.

I have had them for weeks and told no one. Going to start sharing my tests here.

Local AI folks, what is the first model I have to run on this?

[Batuhan Ayci @batuhanayci]

@garrytan Can she pass three fingers on the face test?

[Batuhan Ayci @batuhanayci]

@MiaAI_lab @volatilemarkts Good work! 🤯

[Batuhan Ayci @batuhanayci]

@VictorTaelin The question is do you really need a frontier level model

[Batuhan Ayci @batuhanayci]

@loktar00 How does your electric bill look like?

[Batuhan Ayci @batuhanayci]

@ItsmeAjayKV @ideogram_ai So many models are released in a very short period, it is impossible to have independent analysis done... I am curious when it will land on artificial analysis

[Batuhan Ayci @batuhanayci]

@MikeBradleyAI Do you think it is better than GLM-5.3-Flash?

[Batuhan Ayci @batuhanayci]

@ashxhart Probably your mess, you skyrocketed DGX Spark prices after releasing Tensorfold 🤯

[Batuhan Ayci @batuhanayci]

@TechMDAI I love working on TP2...

[Batuhan Ayci @batuhanayci]

@loktar00 2x dgx sparks are not even 20k now 🤯

[Batuhan Ayci @batuhanayci]

@u1tra_instinct @ashxhart Wish I had four sparks... Great work!

[Batuhan Ayci @batuhanayci]

@Blackfrost_AI Well deserved!

[Batuhan Ayci @batuhanayci]

@ItsmeAjayKV They might distill from Argon to get Gemini-4-pro, who knows

[Batuhan Ayci @batuhanayci]

@kimmonismus Hope to see some quants fitting two DGX Sparks...

[Batuhan Ayci @batuhanayci]

@MiaAI_lab @ArtificialAnlys How many subs do you have, Mia?

[Batuhan Ayci @batuhanayci]

@jun_song Despite unofficial, artificial analysis results are promising