Ashkan Kazemi

AI researcher @Cisco building scalable simulated worlds where RL agents can learn, fail, and improve before real-world deployment

PhD from @umichcse

[Ashkan Kazemi @ashkankazemi]

GPT-5 high reasoning is incredible for writing complicated code but ask for a script rerun and there is non-trivial chance it will overthink and mess it up. Another sign that determining "how much thinking" is necessary for a problem might not be an easy goal for generalist AI.

[Jabez Magomere @jabez_magomere]

This work was a collaboration with @meedan and amazing co-authors:
@computermacgyve @emanuelelamalfa @iperboreo_ @ashkankazemi @ManuelTonneau

📄 Read the paper: https://t.co/DVQHInLBn8

[Dorian Quelle @dorian_quelle]

🚨 New pre-print! 🚨
 “Why Academics Are Leaving Twitter for Bluesky”, joint with  @Frederic_Denker @Prashant_Garg_ @BovetAlexandre We track 300k scholars and find that 18 % have already switched. In this 🧵 who moved why, how contagion works, and what sustains engagement.

[Ashkan Kazemi @ashkankazemi]

@YulinYuResearch @UofAInfoSci @uarizona @UzziLeadership @KelloggSchool @NorthwesternU @MSFTResearch Congrats Yulin, best of luck at University of Arizona!

[Ashkan Kazemi @ashkankazemi]

@baobaoyaobaobao Congratulations Dr. Bao🥳
Excited for you in your next chapter!

[Ashkan Kazemi @ashkankazemi]

@niloofar_mire Congrats Niloofar, very well deserved🥳

also that's a gorgeous shot of Dolores🤩did not know you're based in SF these days, would love to connect!

[Ashkan Kazemi @ashkankazemi]

@ShiboChenTech Congratulations Shibo!

[Ashkan Kazemi @ashkankazemi]

@Iranistics_UT https://t.co/GUWOTweQ8M

[Ahmed Moubtahij @ahmed_moubtahij]

@lateinteraction I lack the bandwidth to work on this atm, but I have a PageRank implementation if anyone wants to hack it: https://t.co/9P4g3MJPqQ

It extends Biased TextRank (@ashkankazemi et al., 2020) with parametric biases (e.g. query formulations, sentiment) and reductions.

[Ashkan Kazemi @ashkankazemi]

@rasoulipour خیلی پروژه‌ی جالبیه، دست مریزاد
احتمالا بتونیم با هوش مصنوعی برای داستان‌ها تصویرسازی هم بکنیم

[Shiri Dori-Hacohen♿️🧠✡️🦓 (🚫🥄s). 🟣🎗️🏴 @ShirKi]

Excited to have our recent work profiled on @meedan's blog! See RT'd🧵 for how we use #LLMs to generate synthetic data and improve local models in multilingual settings.

[Ashkan Kazemi @ashkankazemi]

also a big thank you to @meedan's comms team (Matt and Kipp) for helping me explain my ideas more clearly and designing very cutesy visualizations❣️

Link to blog post: https://t.co/1yH0HkROPk

SynDy paper: https://t.co/AiqycPnhwt

[Ashkan Kazemi @ashkankazemi]

Check out my blog post on @meedan's recent work on using LLMs for running a multi language classification service called ClassyCat🐈‍⬛

details below🧵

[Meedan @meedan]

💻 At Meedan, we’re focused on leveraging emerging technology, such as generative AI, to promote efficient, ethical, and culturally informed decision-making. Find out how we’re putting it to work with the development of our new internal tool ClassyCat. 🐈‍⬛

https://t.co/8s075PaskT

[Ashkan Kazemi @ashkankazemi]

interested in our research efforts at @meedan helping global journalism? feel free to reach out, or leave a comment here!

big shout out to my collaborators on this effort: my manager and mentor @computermacgyve, and @ShirKi and the UConn team for taking this off ground w/ SynDy