About
I like to build products driven by AI research that is indistinguishable from magic. I…
Services
Articles by Delip
-
Detecting Forged/Synthetic Content
Detecting Forged/Synthetic Content
I enjoyed speaking at the National Academies of Sciences today on detection of synthetic content like #deepfakes and…
1 Comment
Activity
2K followers
-
Delip Rao shared thisWhat does a 15% overhead cap really mean for research funding? In my latest post, I break down the numbers behind NIH’s cap and its impact on academia and innovation. Dive in! ⬇️ https://lnkd.in/eB-MzNBk
-
Delip Rao reposted thisDelip Rao reposted this‘WithdrarXiv’ database of 14,000 retracted preprints launches‘WithdrarXiv’ database of 14,000 retracted preprints launches
-
Delip Rao shared thistook some time to breakdown OpenAI's O1 press release to offer my unfiltered thoughts. More on substack: https://lnkd.in/exmqBD-s
-
Delip Rao reposted thisHere's a nice profile by Penn Engineering about our research into detecting LLM-generated text. My PhD student Liam created a new benchmark dataset and leaderboard for the task, which he's presenting at the ACL conference this week.Delip Rao reposted thisLarge language models aid in human creativity and project efficiency, but their power can be misused or lead to misinformation. In a recent article, Chris Callison-Burch, Professor in Computer and Information Science, and Ph.D. student Liam Dugan (pictured below) introduce the Robust AI Detector (RAID) dataset for evaluating AI text detection. The RAID dataset helps identify the limitations of current detectors and paves the way for developing more reliable tools. As Callison-Burch says, “It’s an arms race,” highlighting the need for advanced technology to keep pace with evolving AI. Learn how their research shapes the future of AI detection and promotes transparency in the field. https://bit.ly/3M0ae0C Penn Computer and Information Science
-
Delip Rao reposted thisDelip Rao reposted thisI am excited to share that we are hiring for Engineering roles for the Child Safety team in São Paulo, Brazil! Join us in our mission to fight online child sexual abuse and exploitation and keeping the internet safe. To learn more about the team, please visit https://lnkd.in/gRypS4vn For role specific positions please see: Software Engineer II, Child Safety Engineering https://lnkd.in/gUjeAuW9 Software Engineer III, Child Safety Engineering https://lnkd.in/gcdnRzAY Engineering Manager, Child Safety Engineering https://lnkd.in/gUrE3MBU
-
Delip Rao reposted thisDelip Rao reposted this🚀 Become a sponsor at EACL 2024 🚀 Everyone wants to do AI but where do you find the AI experts? Join us at EACL 2024 as a sponsor and engage with the AI and NLP community to recruit the AI talent you need! Besides the regular bronze to diamond sponsorship packages, we are looking for D&I champion and ally sponsors. D&I sponsorship will fund travel expenses, accommodation, and conference registration for individuals from economically disadvantaged countries. This will enable 12 authors from the Global South to attend, fostering diversity in the AI community. Why sponsor? 1️⃣ Impactful Contribution 2️⃣ Recruitment Goldmine 3️⃣ Positive Brand Image Find all the sponsorship brochure on the EACL website or DM me by 18 Feb: https://lnkd.in/gZ_JrCgA Don't miss this chance to make a difference and connect with top AI talent. Contact us to learn more about sponsorship opportunities! Let's build a more inclusive AI future together. #EACL2024 #DiversityandInclusion #AICommunity
-
Delip Rao reposted thisDelip Rao reposted thisI'm helping to organize this year's International Conference on Computational Social Science (https://ic2s2-2024.org/), which will be held here at the University of Pennsylvania in Philadelphia from July 18-20, 2024. We've got a great line up of invited speakers including: * Elizabeth Bruch, University of Michigan * Matthew Gentzkow, Stanford University * Dirk Helbing, ETH Zurich * Dan Jurafsky, Stanford University * Amy Orben, University of Cambridge * Desmond Patton, University of Pennsylvania * Talal Rahwan, NYU Abu Dhabi If you'd like to participate in the conference, I encourage you to sumit an abstract! Important dates Abstract submission deadline: February 24, 2024 Notification of acceptance: April 15, 2024 Early-bird registration deadline: May 8, 2024 Conference days: July 18-20, 2024 Call for abstracts The International Conference on Computational Social Science (IC2S2) is the premier conference bringing together researchers from different disciplines interested in using computational and data-intensive methods to solve problems relevant to society. IC2S2 hosts academics and practitioners in computational science, complexity, network science, and social science, and provides a platform for new research in the field of computational social science. Submissions website: https://lnkd.in/eGXnuurq Types of contributions: The accepted contributions will be selected for one of the following presentations: (i) a lightning talk (~6 mins) in a plenary session, (ii) an oral presentation in parallel tracks, or (iii) a poster presentation session. Lightning talks will be preferentially assigned to those requesting this form of presentation at submission and to early career researchers. In order to be included in the program, at least one of the authors must register for the conference by the early-bird registration deadline. No published proceedings: Submissions will be non-archival, and thus the presented work can be already published, in preparation for publication elsewhere, or ongoing research. Abstracts will be reviewed by multiple members of a Program Committee composed of experts in computational social science. Topics All topics in computational social science! Check out our website :) https://ic2s2-2024.org/ Enquiries For any questions, please write to: ic2s2.2024@gmail.com
-
Delip Rao shared thisI am writing a series of posts situating The New York Times copyright case against OpenAI in hard technical realities (science) to further an informed discussion of these topics. I just published the first part, and I would very much like your feedback. This case reignited generative AI’s impact on independent writers and artists, and copyright is more complex than copying somebody’s text or image without permission. Often, the discussions around copyright are passioned and are centered around expressions of (justified) anger and blame, ideas on what we would like to happen that are disconnected from technical realities. My hope in writing this is that by knowing the technological forces we are interacting with, we can better shape public conversations, laws, and policies. I would love your feedback on this series, starting with this first articleThe NYT vs. OpenAI Copyright Case: A Technical Introduction – Part IThe NYT vs. OpenAI Copyright Case: A Technical Introduction – Part I
-
Delip Rao reposted thisDelip Rao reposted thisIf I am going to amplify the voices who say it is going to all crash and burn, I should in fairness amplify the voices who say the opposite. Delip Rao explains why he thinks LLMs are here to stay, at prices at or below the cost of electricity.
-
Delip Rao reacted on thisDelip Rao reacted on thisBen Church and his thesis advisor (Ravi Vakil) at Stanford graduation.
-
Delip Rao reacted on thisDelip Rao reacted on thisIt has been a tradition for new members of the American Academy of Arts and Sciences to acknowledge their acceptance of membership with a letter. This is the letter I sent to the Academy last month to express my gratitude for this honor. I give all the credit to my family and the values and resilience they taught me during times of hardship, and to the students and collaborators literally from all corners of the world and all walks of life, from whom I have learned so much! Thank you! ❤️
-
Delip Rao liked thisDelip Rao liked this🚀 Such an impressive lineup of projects to conclude our IR & Web Search class — the best reward after a semester of teaching. Nearly 150 students presented their semester-long work on using IR/AI/NLP to address socially relevant problems. Some highlights included autocratic signals in political speeches, models of homelessness policy effectiveness, ranking algorithms for women’s healthcare providers, language-based sentiment bias, support for fair political voting, and so much more. Kudos to all the students for their impactful work! 👏
-
Delip Rao liked thisDelip Rao liked thisOkay y'all, it's official. I'm a co-organizer of the NeurIPS 2026 Atlanta Satellite site. See you in December
-
Delip Rao reacted on thisDelip Rao reacted on thisJust got back from Skyroot's campus in Hyderabad, and I've got to say, visiting your launch provider is different from just signing documents. Rishi and I have been buried in subsystems for months now. Zenith-1 is a 1U CubeSat designed to measure how cosmic radiation messes with SRAM in orbit. We're trying to correlate transistor size with Single Event Upset rates in LEO. It's ambitious for a Class 11 project from Pune, and we've got Skyroot taking us to 450 km. The thing nobody really tells you: even after you've designed everything, you don't actually know how your satellite fits into the mission until you talk to the people launching it. Aniruddh Sir walked us through the deployer interface, the mechanical constraints, the timeline. Suddenly all those abstract subsystem decisions (the power budget, the heat dissipation, the form factor) became real. They mattered. We also got into the weeds on environmental testing: how cold is too cold during outgassing, whether our Raspberry Pi 5 can actually survive 40+ thermal cycles, what "flight ready" actually looks like when you're a student team. The vibe was less "you kids built a toy" and more "this is what space engineering actually requires." That matters to us. Looking forward to integration phase. And to November.
-
Delip Rao reacted on thisDelip Rao reacted on thisTwo Penn Engineering faculty members, Chris Callison-Burch and Gad Allon, have been named recipients of the 2026 Lindback Awards for Distinguished Teaching, among the University’s highest honors for faculty. Callison-Burch, Professor in CIS and Program Director of Penn Engineering’s online Master’s in Artificial Intelligence, is a leading researcher in AI and natural language processing. His teaching brings cutting-edge work in machine translation and large language models directly into the classroom across both on-campus and online programs. Allon, the Jeffrey A. Keswin Professor in Operations, Information and Decisions at Wharton with a secondary appointment in ESE, teaches operations strategy and scaling. His courses are known for connecting business and engineering perspectives, grounded in research on how organizations design and scale complex systems. Established in 1961 with support from the Christian R. and Mary F. Lindback Foundation, the Lindback Awards recognize excellence in teaching across the University. Congratulations, Chris and Gad! https://bit.ly/4ckMLnT Penn Engineering AI / Penn Computer and Information Science / Electrical and Systems Engineering at Penn / The Wharton School
-
Delip Rao reacted on thisDelip Rao reacted on thisLast Friday, March 6 2026, I retired from 18+ years at Google Research and Google DeepMind working to bring AI for language understanding and recommendations to Google products. What we accomplished, thanks to my many wonderful colleagues along the way, was far beyond what I had hoped when I came to Google in 2008 motivated by the belief that ML was ready to drive advances in language and user understanding. This Monday, I rekindle another long-standing interest, joining Inceptive as their Chief AI Scientist, where I will build on what I learned over the last two decades to advance biological foundation models that enable faster biomedical discovery.
Experience
-
MangoMind AI
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Education
-
The Johns Hopkins University
-
-
Activities and Societies: Association for Computational Linguistics
My thesis is on large scale graph based learning algorithms and their applications to natural language processing problems.
-
-
-
-
-
-
-
-
-
Licenses & Certifications
Publications
Recommendations received
1 person has recommended Delip
Join now to viewOther similar profiles
Explore more posts
-
Harvey Yorke
Valyu DeepResearch (fast mode) is #1 on ScholarQA - the Nature-published benchmark for AI scientific literature reviews. 29% ahead of OpenScholar. Beating Perplexity Pro and PaperQA2. ScholarQA is the leading benchmark for AI-generated scientific literature reviews, published in Nature Feb 2026. Built by researchers at UW, Allen AI, Meta, and CMU. This was our lowest cost system, fast mode. This mode uses ~20x less compute than our Heavy mode, and ~40x less than Max mode. How: Valyu is the search API for AI Knowledge work, with full-text multimodal indexing across PubMed, arXiv, bio/medRxiv, clinical trials, CheMBL, DailyMed, and more. Going beyond full-text, and with powerful semantic search, we achieved SoTA results, with 2x more citations per answer than OpenScholar, with real DOIs and authors attached. Full benchmark writeup and code in the comments.
14 Comments -
Tobalo T.
DeepSeek recently released an OCR model, marking another significant milestone in AI engineering. This remarkable capability involves utilizing the visual modality as an efficient compression medium for textual information, addressing the challenges posed by long-context scenarios in LLMs. DeepSeek is also leveraging this model to augment their training data for their models. They are advocating for the use of generating training datasets with internal and/or institutional documents (dark data) to enhance the model’s performance. DeepSeek-OCR has the remarkable ability to generate training data for LLMs/VLMs at an impressive scale of over 200,000 pages per day, utilizing a single A100-40G GPU. This information compression capability is highly valuable for context engineering document processing workflows.
-
Bob Hayes, PhD
The Scaling Fallacy: Bigger LLM won't lead to #AGI "An honest framing could be: AGI might emerge from complex AI systems that include LLMs as one component. But perhaps that’s a much less exciting fundraising pitch than: Scale LLMs → AGI!" ~ Wendy Wee https://lnkd.in/gVpdVYRC #GenerativeAI
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content