Aishik Ghosh
@aishikghosh.bsky.social
Assistant Professor of AI in School of Physics @ GT
Fundamental Physics ∩ AI
AI Policy
AI for hypothesis generation, simulation, inference in particle & astro
High-dimensional statistics, neural inference, uncertainty quantification
Fundamental Physics ∩ AI
AI Policy
AI for hypothesis generation, simulation, inference in particle & astro
High-dimensional statistics, neural inference, uncertainty quantification
Our two new research areas are now live!
November 18, 2025 at 3:36 PM
Our two new research areas are now live!
📢 𝐆𝐫𝐚𝐝 𝐚𝐝𝐦𝐢𝐬𝐬𝐢𝐨𝐧𝐬 𝐨𝐩𝐞𝐧, 𝐞𝐚𝐫𝐥𝐢𝐞𝐫 𝐝𝐞𝐚𝐝𝐥𝐢𝐧𝐞 𝐭𝐡𝐢𝐬 𝐲𝐞𝐚𝐫!
Physics at 𝐆𝐞𝐨𝐫𝐠𝐢𝐚 𝐓𝐞𝐜𝐡 is accepting applications.
𝐌𝐲 𝐠𝐫𝐨𝐮𝐩 𝐨𝐧 𝐀𝐈 ∩ 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 𝐢𝐬 𝐠𝐫𝐨𝐰𝐢𝐧𝐠! So if you know students interested in 𝐟𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 𝐨𝐫 𝐚𝐬𝐭𝐫𝐨, please share this.
🗓️ Deadline: 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫 𝟏𝟓
🔗 physics.gatech.edu/academics/gr...
Physics at 𝐆𝐞𝐨𝐫𝐠𝐢𝐚 𝐓𝐞𝐜𝐡 is accepting applications.
𝐌𝐲 𝐠𝐫𝐨𝐮𝐩 𝐨𝐧 𝐀𝐈 ∩ 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 𝐢𝐬 𝐠𝐫𝐨𝐰𝐢𝐧𝐠! So if you know students interested in 𝐟𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 𝐨𝐫 𝐚𝐬𝐭𝐫𝐨, please share this.
🗓️ Deadline: 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫 𝟏𝟓
🔗 physics.gatech.edu/academics/gr...
Admissions Information
Prospective physics graduate students are welcome to contact faculty members to inquire about their research program. However, individual faculty members cannot accept students into the graduate progr...
physics.gatech.edu
November 11, 2025 at 4:55 PM
📢 𝐆𝐫𝐚𝐝 𝐚𝐝𝐦𝐢𝐬𝐬𝐢𝐨𝐧𝐬 𝐨𝐩𝐞𝐧, 𝐞𝐚𝐫𝐥𝐢𝐞𝐫 𝐝𝐞𝐚𝐝𝐥𝐢𝐧𝐞 𝐭𝐡𝐢𝐬 𝐲𝐞𝐚𝐫!
Physics at 𝐆𝐞𝐨𝐫𝐠𝐢𝐚 𝐓𝐞𝐜𝐡 is accepting applications.
𝐌𝐲 𝐠𝐫𝐨𝐮𝐩 𝐨𝐧 𝐀𝐈 ∩ 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 𝐢𝐬 𝐠𝐫𝐨𝐰𝐢𝐧𝐠! So if you know students interested in 𝐟𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 𝐨𝐫 𝐚𝐬𝐭𝐫𝐨, please share this.
🗓️ Deadline: 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫 𝟏𝟓
🔗 physics.gatech.edu/academics/gr...
Physics at 𝐆𝐞𝐨𝐫𝐠𝐢𝐚 𝐓𝐞𝐜𝐡 is accepting applications.
𝐌𝐲 𝐠𝐫𝐨𝐮𝐩 𝐨𝐧 𝐀𝐈 ∩ 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 𝐢𝐬 𝐠𝐫𝐨𝐰𝐢𝐧𝐠! So if you know students interested in 𝐟𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 𝐨𝐫 𝐚𝐬𝐭𝐫𝐨, please share this.
🗓️ Deadline: 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫 𝟏𝟓
🔗 physics.gatech.edu/academics/gr...
📢𝐉𝐨𝐛 𝐚𝐥𝐞𝐫𝐭: Postdoc position(s) in 𝐀𝐈 𝐟𝐨𝐫 𝐟𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 focusing on either 𝐬𝐭𝐚𝐭𝐢𝐬𝐭𝐢𝐜𝐚𝐥 𝐢𝐧𝐟𝐞𝐫𝐞𝐧𝐜𝐞 or 𝐀𝐈-𝐚𝐬𝐬𝐢𝐬𝐭𝐞𝐝 𝐭𝐡𝐞𝐨𝐫𝐲.
🏙️Join my new group at 𝐆𝐞𝐨𝐫𝐠𝐢𝐚 𝐓𝐞𝐜𝐡, based in 𝐰𝐨𝐧𝐝𝐞𝐫𝐟𝐮𝐥 𝐌𝐢𝐝𝐭𝐨𝐰𝐧 𝐀𝐭𝐥𝐚𝐧𝐭𝐚!
🗓️ Reviews begin sooner, deadline: 𝟓𝐭𝐡 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫
💬 Questions: Reach out!
🔗 inspirehep.net/jobs/2962185
🏙️Join my new group at 𝐆𝐞𝐨𝐫𝐠𝐢𝐚 𝐓𝐞𝐜𝐡, based in 𝐰𝐨𝐧𝐝𝐞𝐫𝐟𝐮𝐥 𝐌𝐢𝐝𝐭𝐨𝐰𝐧 𝐀𝐭𝐥𝐚𝐧𝐭𝐚!
🗓️ Reviews begin sooner, deadline: 𝟓𝐭𝐡 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫
💬 Questions: Reach out!
🔗 inspirehep.net/jobs/2962185
INSPIRE
inspirehep.net
November 10, 2025 at 2:21 PM
📢𝐉𝐨𝐛 𝐚𝐥𝐞𝐫𝐭: Postdoc position(s) in 𝐀𝐈 𝐟𝐨𝐫 𝐟𝐮𝐧𝐝𝐚𝐦𝐞𝐧𝐭𝐚𝐥 𝐩𝐡𝐲𝐬𝐢𝐜𝐬 focusing on either 𝐬𝐭𝐚𝐭𝐢𝐬𝐭𝐢𝐜𝐚𝐥 𝐢𝐧𝐟𝐞𝐫𝐞𝐧𝐜𝐞 or 𝐀𝐈-𝐚𝐬𝐬𝐢𝐬𝐭𝐞𝐝 𝐭𝐡𝐞𝐨𝐫𝐲.
🏙️Join my new group at 𝐆𝐞𝐨𝐫𝐠𝐢𝐚 𝐓𝐞𝐜𝐡, based in 𝐰𝐨𝐧𝐝𝐞𝐫𝐟𝐮𝐥 𝐌𝐢𝐝𝐭𝐨𝐰𝐧 𝐀𝐭𝐥𝐚𝐧𝐭𝐚!
🗓️ Reviews begin sooner, deadline: 𝟓𝐭𝐡 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫
💬 Questions: Reach out!
🔗 inspirehep.net/jobs/2962185
🏙️Join my new group at 𝐆𝐞𝐨𝐫𝐠𝐢𝐚 𝐓𝐞𝐜𝐡, based in 𝐰𝐨𝐧𝐝𝐞𝐫𝐟𝐮𝐥 𝐌𝐢𝐝𝐭𝐨𝐰𝐧 𝐀𝐭𝐥𝐚𝐧𝐭𝐚!
🗓️ Reviews begin sooner, deadline: 𝟓𝐭𝐡 𝐃𝐞𝐜𝐞𝐦𝐛𝐞𝐫
💬 Questions: Reach out!
🔗 inspirehep.net/jobs/2962185
Reposted by Aishik Ghosh
🚨 Job Alert
We're hiring our first Postdoc to work with us on our @erc.europa.eu Project on Generative AI for Particle Physics.
Check out the details below and don't hesitate to get in touch or spread the word!
Deadline: Nov 30th
inspirehep.net/jobs/3075448
We're hiring our first Postdoc to work with us on our @erc.europa.eu Project on Generative AI for Particle Physics.
Check out the details below and don't hesitate to get in touch or spread the word!
Deadline: Nov 30th
inspirehep.net/jobs/3075448
LinkedIn
This link will take you to a page that’s not on LinkedIn
lnkd.in
October 31, 2025 at 6:26 PM
🚨 Job Alert
We're hiring our first Postdoc to work with us on our @erc.europa.eu Project on Generative AI for Particle Physics.
Check out the details below and don't hesitate to get in touch or spread the word!
Deadline: Nov 30th
inspirehep.net/jobs/3075448
We're hiring our first Postdoc to work with us on our @erc.europa.eu Project on Generative AI for Particle Physics.
Check out the details below and don't hesitate to get in touch or spread the word!
Deadline: Nov 30th
inspirehep.net/jobs/3075448
Setting up my research group Slack workspace (free version). Has anyone considered Discord / Zulip / something else? @marielpettee.bsky.social @lukasheinrich.com
October 16, 2025 at 3:25 PM
Setting up my research group Slack workspace (free version). Has anyone considered Discord / Zulip / something else? @marielpettee.bsky.social @lukasheinrich.com
Reposted by Aishik Ghosh
Yeah, even crazier- last year for my postdoc ad a full half of the applications legit never made reference to anything my lab does and how my research would connect to theirs. You are literally in the top half of applicants for just TRYING to draw a connection in a sentence in the cover letter!
October 7, 2025 at 11:09 PM
Yeah, even crazier- last year for my postdoc ad a full half of the applications legit never made reference to anything my lab does and how my research would connect to theirs. You are literally in the top half of applicants for just TRYING to draw a connection in a sentence in the cover letter!
Like AI + Physics, which is a well-developed field but we're still only scratching the surface of what we can do!
For folks considering grad school in ML, my advice is to explore programs that mix ML with a domain interest. ML programs are wildly oversubscribed while a lot of the fun right now is in figuring out what you can do with it
September 25, 2025 at 6:43 AM
Like AI + Physics, which is a well-developed field but we're still only scratching the surface of what we can do!
At the launch event for @gtsciences.bsky.social's AI4Science Centre: ai4science.ai.gatech.edu. With colleagues from psychology, math, earth sciences and everything in between. Looking forward to all the cross-pollination within the sciences and also with industry!
August 26, 2025 at 10:12 PM
At the launch event for @gtsciences.bsky.social's AI4Science Centre: ai4science.ai.gatech.edu. With colleagues from psychology, math, earth sciences and everything in between. Looking forward to all the cross-pollination within the sciences and also with industry!
A little comment like one of these makes my week. Teaching and mentorship is easily the most rewarding aspect of academia. Makes it all worth it. Thanks again to all the students who came to ML4FP and interacted with us for a week. Your positive comments mean the word to us!
August 14, 2025 at 4:39 PM
A little comment like one of these makes my week. Teaching and mentorship is easily the most rewarding aspect of academia. Makes it all worth it. Thanks again to all the students who came to ML4FP and interacted with us for a week. Your positive comments mean the word to us!
Today we have the ATLAS, CMS and Neutrino parallel sessions at the ML4FP School. Julia Gonski give us an overview of ML in ATLAS
August 8, 2025 at 4:36 PM
Today we have the ATLAS, CMS and Neutrino parallel sessions at the ML4FP School. Julia Gonski give us an overview of ML in ATLAS
GPT5's explanation for FTS is quite impressive, especially where it gives an intuitive explanation that is not just directly lifted from the paper. Makes it easy to learn a new statistics concept.
chatgpt.com/s/t_6894ece0...
chatgpt.com/s/t_6894ece0...
GPT-5 vs GPT-4
chatgpt.com
August 7, 2025 at 6:21 PM
GPT5's explanation for FTS is quite impressive, especially where it gives an intuitive explanation that is not just directly lifted from the paper. Makes it easy to learn a new statistics concept.
chatgpt.com/s/t_6894ece0...
chatgpt.com/s/t_6894ece0...
And we're off! Yifan Chen welcomes students to the Machine Learning for Fundamental Physics School 2025!
August 4, 2025 at 3:50 PM
And we're off! Yifan Chen welcomes students to the Machine Learning for Fundamental Physics School 2025!
What if you could beat the Likelihood Ratio Test?
A fun and potentially transformational phy-statistics paper with Ann Lee, James Carzon, @rafael-izbicki.bsky.social, luca Masserano and @danielwhiteson.bsky.social
arxiv.org/abs/2507.17831
A fun and potentially transformational phy-statistics paper with Ann Lee, James Carzon, @rafael-izbicki.bsky.social, luca Masserano and @danielwhiteson.bsky.social
arxiv.org/abs/2507.17831
On Focusing Statistical Power for Searches and Measurements in Particle Physics
Particle physics experiments rely on the (generalised) likelihood ratio test (LRT) for searches and measurements, which consist of composite hypothesis tests. However, this test is not guaranteed to b...
arxiv.org
July 28, 2025 at 7:02 AM
What if you could beat the Likelihood Ratio Test?
A fun and potentially transformational phy-statistics paper with Ann Lee, James Carzon, @rafael-izbicki.bsky.social, luca Masserano and @danielwhiteson.bsky.social
arxiv.org/abs/2507.17831
A fun and potentially transformational phy-statistics paper with Ann Lee, James Carzon, @rafael-izbicki.bsky.social, luca Masserano and @danielwhiteson.bsky.social
arxiv.org/abs/2507.17831
Reposted by Aishik Ghosh
Happy this paper is finally on arxiv!
A fun mix of physics (jet substructure) being used to calibrate the performance of ML algorithms (jet taggers) to get proper uncertainties
Now a standard CMS tool for interpreting searches for exotic jets, used in multiple analyses
A fun mix of physics (jet substructure) being used to calibrate the performance of ML algorithms (jet taggers) to get proper uncertainties
Now a standard CMS tool for interpreting searches for exotic jets, used in multiple analyses
#CMSpaper: A method for correcting the substructure of multiprong jets using the Lund jet plane (arXiv:2507.07775) https://arxiv.org/abs/2507.07775 #Jets
July 14, 2025 at 5:24 PM
Happy this paper is finally on arxiv!
A fun mix of physics (jet substructure) being used to calibrate the performance of ML algorithms (jet taggers) to get proper uncertainties
Now a standard CMS tool for interpreting searches for exotic jets, used in multiple analyses
A fun mix of physics (jet substructure) being used to calibrate the performance of ML algorithms (jet taggers) to get proper uncertainties
Now a standard CMS tool for interpreting searches for exotic jets, used in multiple analyses
New paper: Constraining the Higgs self-coupling from the off-shell single Higgs with NSBI. Work with Tae Hyoun Park, Maximilian Griese & Ulrich Haisch.
arxiv.org/abs/2507.02032

arxiv.org/abs/2507.02032

July 4, 2025 at 2:12 AM
New paper: Constraining the Higgs self-coupling from the off-shell single Higgs with NSBI. Work with Tae Hyoun Park, Maximilian Griese & Ulrich Haisch.
arxiv.org/abs/2507.02032

arxiv.org/abs/2507.02032

Reposted by Aishik Ghosh
So happy to see @aishikghosh.bsky.social thrive - he’s been one of the drivers of AI in HEP and I can’t wait to see his lab start up - take note of the postdoc position coming up.
Some happy news — I’ve accepted a tenure-track faculty position in the School of Physics at Georgia Tech! My group will work at the intersection of AI and fundamental physics & astrophysics. GT is making massive investments in these directions, and I’m hiring a postdoc — more soon!
June 24, 2025 at 6:27 AM
So happy to see @aishikghosh.bsky.social thrive - he’s been one of the drivers of AI in HEP and I can’t wait to see his lab start up - take note of the postdoc position coming up.
Some happy news — I’ve accepted a tenure-track faculty position in the School of Physics at Georgia Tech! My group will work at the intersection of AI and fundamental physics & astrophysics. GT is making massive investments in these directions, and I’m hiring a postdoc — more soon!
June 23, 2025 at 6:19 PM
Some happy news — I’ve accepted a tenure-track faculty position in the School of Physics at Georgia Tech! My group will work at the intersection of AI and fundamental physics & astrophysics. GT is making massive investments in these directions, and I’m hiring a postdoc — more soon!
Delighted to see our work featured in the news! Two key players mentioned but unfortunately not explicitly named: discussions with Kyle Cranmer and Johann Brehmer were crucial in the initial stages. Also Jay, Rafael were central in making it a reality in ATLAS.
Our work on simulation-based inference at the LHC is featured in @arstechnica.com! Well done @aishikghosh.bsky.social @sandesarajay.bsky.social 🙌 arstechnica.com/science/2025...
How a grad student got LHC data to play nice with quantum interference
New approach is already having an impact on the experiment’s plans for future work.
arstechnica.com
June 23, 2025 at 5:52 PM
Delighted to see our work featured in the news! Two key players mentioned but unfortunately not explicitly named: discussions with Kyle Cranmer and Johann Brehmer were crucial in the initial stages. Also Jay, Rafael were central in making it a reality in ATLAS.
Reposted by Aishik Ghosh
Our work on simulation-based inference at the LHC is featured in @arstechnica.com! Well done @aishikghosh.bsky.social @sandesarajay.bsky.social 🙌 arstechnica.com/science/2025...
How a grad student got LHC data to play nice with quantum interference
New approach is already having an impact on the experiment’s plans for future work.
arstechnica.com
June 23, 2025 at 1:55 PM
Our work on simulation-based inference at the LHC is featured in @arstechnica.com! Well done @aishikghosh.bsky.social @sandesarajay.bsky.social 🙌 arstechnica.com/science/2025...
Wondered what it would take for AI agents to design and test new theories for fundamental physics, just like a physicist? We show how, an example in neutrino model building! Paper with Jake, Max, Victoria, Jason leading crucial aspects, also
@danielwhiteson.bsky.social
.
arxiv.org/abs/2506.08080
@danielwhiteson.bsky.social
.
arxiv.org/abs/2506.08080
June 11, 2025 at 4:10 PM
Wondered what it would take for AI agents to design and test new theories for fundamental physics, just like a physicist? We show how, an example in neutrino model building! Paper with Jake, Max, Victoria, Jason leading crucial aspects, also
@danielwhiteson.bsky.social
.
arxiv.org/abs/2506.08080
@danielwhiteson.bsky.social
.
arxiv.org/abs/2506.08080
Breakthrough Prize awarded to LHC experiments for the study of Higgs properties (to which I contributed), rare processes, matter-antimatter asymmetry, and many other areas of research
An unexpected surprise. The 2025 Breakthrough Prize in Fundamental Physics honors over 13,000 researchers whose labors have led to the precise description the Higgs mechanism, … breakthroughprize.org/News/91 @CERN
April 5, 2025 at 11:46 PM
Breakthrough Prize awarded to LHC experiments for the study of Higgs properties (to which I contributed), rare processes, matter-antimatter asymmetry, and many other areas of research
Reposted by Aishik Ghosh
New paper!
Learning Broken Symmetries with Approximate Invariance
Led by HIGH SCHOOL STUDENT Seth Nabat, with Aishik_Ghosh, Ed Witkowski and GregorKasieczka.
arxiv.org/abs/2412.18773
Learning Broken Symmetries with Approximate Invariance
Led by HIGH SCHOOL STUDENT Seth Nabat, with Aishik_Ghosh, Ed Witkowski and GregorKasieczka.
arxiv.org/abs/2412.18773
Learning Broken Symmetries with Approximate Invariance
Recognizing symmetries in data allows for significant boosts in neural network training, which is especially important where training data are limited. In many cases, however, the exact underlying sym...
arxiv.org
January 6, 2025 at 4:36 PM
New paper!
Learning Broken Symmetries with Approximate Invariance
Led by HIGH SCHOOL STUDENT Seth Nabat, with Aishik_Ghosh, Ed Witkowski and GregorKasieczka.
arxiv.org/abs/2412.18773
Learning Broken Symmetries with Approximate Invariance
Led by HIGH SCHOOL STUDENT Seth Nabat, with Aishik_Ghosh, Ed Witkowski and GregorKasieczka.
arxiv.org/abs/2412.18773
Reposted by Aishik Ghosh
We show that a hybrid (conserving+breaking) architecture combines the advantages of symmetry-conserving networks (fast learning) and breaking ones (higher maximum performance).
Great work by Seth with @aishikghosh.bsky.social, Ed, @danielwhiteson.bsky.social
Full paper: arxiv.org/abs/2412.18773
Great work by Seth with @aishikghosh.bsky.social, Ed, @danielwhiteson.bsky.social
Full paper: arxiv.org/abs/2412.18773
January 6, 2025 at 8:44 AM
We show that a hybrid (conserving+breaking) architecture combines the advantages of symmetry-conserving networks (fast learning) and breaking ones (higher maximum performance).
Great work by Seth with @aishikghosh.bsky.social, Ed, @danielwhiteson.bsky.social
Full paper: arxiv.org/abs/2412.18773
Great work by Seth with @aishikghosh.bsky.social, Ed, @danielwhiteson.bsky.social
Full paper: arxiv.org/abs/2412.18773
Reposted by Aishik Ghosh
Now(ish) on arxiv: Learning Broken Symmetries with Approximate Invariance
Invariances (or, as physicists call them: symmetries) of the data can be baked into a ML model to improve performance or data efficiency.
However, in reality, these symmetries are often broken.
How to deal with that?
Invariances (or, as physicists call them: symmetries) of the data can be baked into a ML model to improve performance or data efficiency.
However, in reality, these symmetries are often broken.
How to deal with that?
January 6, 2025 at 8:44 AM
Now(ish) on arxiv: Learning Broken Symmetries with Approximate Invariance
Invariances (or, as physicists call them: symmetries) of the data can be baked into a ML model to improve performance or data efficiency.
However, in reality, these symmetries are often broken.
How to deal with that?
Invariances (or, as physicists call them: symmetries) of the data can be baked into a ML model to improve performance or data efficiency.
However, in reality, these symmetries are often broken.
How to deal with that?
Reposted by Aishik Ghosh
DSI Research Scientist @matthewfeickert.com focuses on data science applications and #opensource software for particle physics. Read his story: ospo.wisc.edu/community/pr...
December 19, 2024 at 8:39 PM
DSI Research Scientist @matthewfeickert.com focuses on data science applications and #opensource software for particle physics. Read his story: ospo.wisc.edu/community/pr...