Mohd Zaki

Postdoctoral Researcher @JohnsHopkins
Prime Minister's Research Fellow, Ph.D. @iitdelhi @m3rgiitd
B. Tech. @MNNITAL
Research: Materials Discovery, NLP, AI

[Mohd Zaki @zaki_here_]

Hello World,
This is MatSciBERT, material science domain language model.
Thank you Tanishq Gupta (@tanishqg5325) , N. M. Anoop Krishnan (@anoopnm007), and Mausam (@mishumausam) for constant support and guidance.

[arXiv CS-CL @arxiv_cscl]

MatSciBERT: A Materials Domain Language Model for Text Mining and Information Extraction https://t.co/6l3LGVtCiM

[Mohit Bansal @mohitban47]

🎉 Excited that our NSF AI Institute for Engaged Learning has been renewed for another 5 years 🙏

Looking forward to leading the core AI efforts on 3D world models for multimodal narrative environments and AI that adapts to student needs -- and collaborating with our wonderful partners (top experts in education and learning sciences) to produce transformative advances in our understanding of how students learn and fundamentally improve the learning outcomes for K-12 students!

🚨 We'll be hiring postdocs+ phd students etc. for the next 5 years for this, so apply+ping with any questions!

[prathosh A P @prathoshap]

Tenured, with an early promotion at @iiscbangalore . Thankful to everyone who helped me along this journey.

When I heard the news, there was a sense of emptiness. My philosophical mind started pondering, “Oh ya, it’s not the destination, it’s the journey.” Then came a Zen saying, “It’s neither the destination nor the journey; it’s the company.”

The Mahabharata is not about five princes ruling a kingdom. It is a tale of self-doubt, denial, struggles, war, and Dharma. What makes it an epic is also the company it keeps, from Vedavyasa, Vidura, Bhishma and Drona to the sages and, of course, Yogeshwara Krishna himself. The conversations, dilemmas, and struggle to uphold Dharma are the crux, not the plot. शिक्षा ग्राह्या, न कथा कथञ्चन।

So I thought I would write a little self-note to look back and thank some of the wonderful souls who built me, brick by brick.

My parents gave me the freedom to be myself and never doubted me when I chose unconventional paths.

There was an old man, Sheshagiri Rao, who sculpted me inch by inch. From differential equations to Nyaya and Vedanta, he taught me for 25 long years. More importantly, he taught me life. He was my friend, philosopher, and guide. Sadly, he left his body last year and is not here to witness the sapling he planted and nurtured bearing fruit. He is the reason I am what I am today.

Prof. @YoganandaC85447 not only taught me mathematics during my undergraduate days, but also inspired me to take up this profession. Prof. Ramarishnanan AG and Dr. TV Ananthapadmanabha introduced me to academic research, while my teachers, Profs. PS Sastry, Chandra Murthy, and Chandrasekar Seelamantula instilled mathematical rigour into my thinking.

During my industry stints, I was fortunate to learn from some exceptional engineering minds, including @ManishGuptaMG1 , @geethamhp , Lalith Meshta Sanjay Bharadwaj. They taught me to bridge the gap between academic ideas and real-world practice.

At @iitdelhi, my faculty colleagues treated me as family and helped me navigate academia. I continue to cherish the mentorship, collaboration, and friendships I found there. @ramgopal_rao

When I moved back to South India, the leadership at IISc welcomed me as a faculty member. And the last five years at @iiscbangalore have been nothing short of a fabulous adventure. From starting new courses to taking leave to start a company, I have been given extraordinary freedom and support. I cannot imagine a better academic environment.

I owe much of my research to my PhD students, Arnab, Ayush, Amit, Oshin, @backpropogator , Subhodip, Pavan, Shubham, and Karn. I have had the joy of watching them transition from young enthusiasts to collaborators and peers.

I am equally grateful to the thousands of students I have taught, both in-class and online. They have given me a unique opportunity to learn every time I walk into a classroom. Special thanks to @chandanjayaram , my right-hand henchman in all my MOOC efforts.

I have also been fortunate to have fantastic collaborators from industry who taught me how to translate mathematics into meaningful applications. And fortune favoured me in entrepreneurship too. Two of my best students, @Maxaravind and Vinay, are now my cofounders at @LatentForceHQ . Our investors @IdeaspringCap and Yali Capital, led by Nagananda and Ganapathy Subramanian, took a chance on a professor trying to build a deep-tech company.

There are many more mentors, colleagues, collaborators, and well-wishers whose love and blessings have shaped my journey. The list is by no means exhaustive, and my apologies to those I have missed.

After writing all this, I feel more and more like I have been a निमित्तमात्रं. From being a small-town kid who didn’t even know that this world existed, to being where I am today, it all feels quite miraculous.

One thing I tell my young friends is: If I can be here, anyone can too.

Tenure, they say, is freedom. With increased freedom comes increased responsibility. I wish and pray that God gives me enough courage, strength, and will to use that freedom responsibly and tread the path of Dharma.

With that note, I shall stop typing. Yes, I typed every word of this without an LLM. And seek your blessings.

Prathosh.

ईशावास्यमिदं सर्वं यत्किञ्च जगत्यां जगत्।
तेन त्यक्तेन भुञ्जीथा मा गृधः कस्यस्विद्धनम्॥
कुर्वन्नेवेह कर्माणि जिजीविषेच्छतं समाः।
एवं त्वयि नान्यथेतोऽस्ति न कर्म लिप्यते नरे॥

[Parul Pandey @pandeyparul]

I used @claudeai Opus 5.5 to improve a project I made last year and I really like how it has turned out.

Last year, I built a small project to make geography more interactive for kids. Thanks to opus 5.5, it has been turned into a better version.

❓How it works:

• Pick a place and watch a plane travel across the globe to get there. You can see its route and the distance covered as it flies. Kids especially like this part because it gives them the feeling that they are actually travelling somewhere, right up to the plane touching down 🛬

• Once you land, you can explore the place in photorealistic 3D, move around, zoom in, and look at it from different angles.

• And if you answer the quiz correctly, you collect a passport stamp for that place.📘

Let me know if you want to try it.

[N M Anoop Krishnan @anoopnm007]

A note for students: acceptances and rejections are both part of research. We had our share of rejections this cycle too, as in every cycle before and ones to come. What stays is the work and what you learn while doing it. Enjoy the good days, and don't let the other slow you.

[N M Anoop Krishnan @anoopnm007]

The ICLR deadline has passed, the group has slept (mostly), so it's time to share some news: four papers from M3RG and our collaborators have been accepted at NeurIPS 2026. 🎉
@iitdelhi
@ANRFIndia
@NeurIPSConf

[Daniel Khashabi 🕊️ @DanielKhashabi]

Congrats to Sophia and the team on their NeurIPS acceptance! 🎉

Actonomy taxonomy gives us a shared vocabulary for describing agent behavior!

[Daniel Khashabi 🕊️ @DanielKhashabi]

Congrats to Tianjian on his "SelfCompact" paper being accepted to NeurIPS! 🎉

As agents work through *long* tasks, their context fills with old mistakes and stale information. SelfCompact lets the model decide *when* to summarize its own work, preserving useful findings while clearing room to keep going.

[Mohd Zaki @zaki_here_]

@taylordsparks I could have met you if I'd know you're in DC. Until next time!

[Mohd Zaki @zaki_here_]

@jainpapers Excellent opportunity!

[Jain Research Papers @jainpapers]

Hi all,

Postdoc position open on DOE Genesis and ARPA-e projects to accelerate materials design for catalysis, ionic conductors, and general materials synthesis!

In interested, please apply here:
https://t.co/pmyEDerB3f

[Daniel Khashabi 🕊️ @DanielKhashabi]

Congrats to Tianjian on his DARLING paper being accepted to NeurIPS! 🎉

Post-training can make language models more repetitive. DARLING tackles this by rewarding responses that are both high quality and meaningfully different from one another.

[prathosh A P @prathoshap]

2 (out of 3 submissions) accepted at Neurips 2026 @NeurIPSConf from our lab at IISc @iiscbangalore. We have a high hit rate at all A* ML conferences (ICML, ICLR, AAAI). Papers on Training efficiency and graph Mamba (SSMs). Details to follow soon.

Congratulations to my students.

[Homa Majd @homamajd]

I’m excited to share that we are looking for a postdoctoral researcher to join my lab at Yale.

If you, or someone you know, might be interested, please feel free to reach out or share this post.

https://t.co/rllwrIdQmm

[Gael Varoquaux 🦋 @GaelVaroquaux]

🧑‍💻🧑‍🏫 I'm recruiting a post-doc to work on Tabular Foundation Models, one of the hotest topics in AI, where we are at the leading edge.

This is an opportunity to develop the next-level tabular AI, blending deep learning and tables.

[C-DAC India @cdacindia]

India strengthens its global supercomputing footprint with three entries in the latest #TOP500 rankings, including PARAM Rudra and AIRAWAT.

Watch this insightful breakdown by Front Page | @Analyticsindiam: https://t.co/Pb95dguGUE

Developed under the National Supercomputing Mission (#NSM) and implemented by @cdacindia under the aegis of @goi_meity and @IndiaDST, these systems are accelerating breakthroughs in artificial intelligence, scientific research, engineering simulations, weather forecasting and data-intensive applications.

#PARAM #PARAMRudra #HPC #Atmanirbharbharat #Highperformancecomputing #TOP500 #NationalSupercomputingMission
#NationalSupercomputingMission

[William Jurayj @ COLM 2026 @williamjurayj]

I did my taxes this week and like every year, it was a pain. The IRS estimates that the average American spends $270 and 13 hours filing. Wouldn't it be great if language models could help?

[Mohd Zaki @zaki_here_]

@hannahglz25 Congratulations!

[Hannah Gonzalez Collison @hannahglz25]

Very honored to have received the best poster award at the Women in Data Science and AI Symposium at Johns Hopkins!

[célina @hanouticelina]

If you like Claude Code or Codex, you should seriously consider running Agents locally as well! The latest small models (like Qwen 3.5) made this a real before/after moment - and the gap keeps closing. Local coding agents are faster, with more reliable tool calling capabilities, still private, and cost $0 in API bills.

We made it super easy for you to run a local agent with the 𝚊𝚐𝚎𝚗𝚝𝚜 Hugging Face CLI extension - a one-liner that uses 𝚕𝚕𝚖𝚏𝚒𝚝 to detect your hardware and pick the best model and quant, spins up a 𝚕𝚕𝚊𝚖𝚊.𝚌𝚙𝚙 server, and launches Pi (the agent behind OpenClaw 🦞).
One command to find what runs on your hardware and go straight to a working local coding agent!

You should give it a try! 👇

[Mohd Zaki @zaki_here_]

@mohitban47 A similar beautiful library is in Baltimore, George Peabody Library at Johns Hopkins!
Do visit when visiting Hopkins