We are hiring a Ph.D. intern @ NVIDIA for research on multi-modal generative models, reinforcement learning, self-supervised representations, and 3D perception for video conferencing. If you have experience leading insightful contributions on any of these topics, DM / email me.
Today is my first day
@NVIDIA
as a Research Scientist for Human Performance and Experience.
Getting to work with + learn from many world experts to advance our understanding & application of AI is truly a dream come true. I couldn’t have asked for a better place to pursue it! 💚
We are hiring a Ph.D. Intern @ NVIDIA, to work on generative models, visual perception, and quality of digital avatars. Ideal candidate is someone with a track record of leading impactful work. Contact
@RuthRosenholtz
or me if you think you would be a good fit. Details in thread.
📢 We are hiring a summer Ph.D. intern for the Human Performance & Experience team @ NVIDIA Research! Topics include, but are not limited to: deep perceptual quality metrics for visual media, modeling aspects of human vision, & robust representation learning. [1/2]
Incredibly proud of my friend, Abhishek Badki, and my Nvidia mentor,
@0razio
Gallo, for receiving the best student paper Honorable Mention at
#CVPR2021
for their work on Binary TTC: A Temporal Geofence for Autonomous Navigation
Big congrats to all authors!
I’m happy to share that from today I’m starting a new position as a Research Intern at
@nvidia
.
I will be working with
@ekta_prashnani
, Joohwan Kim and Iuri Frosio.
This past month I had the opportunity to mentor 3 exceptional high-schoolers for their intro-to-ML-research projects. Yesterday they gave their final talks & it makes me marvel at how far they have come! A good reminder of how much one can learn if they stay curious & persistent.
As I complete my first year of being a research scientist at NVIDIA, I cannot help but echo the sentiment
@twominutepapers
conveyed in this video: “What a time to be alive!” 😃
Tomorrow at CVPR, I am presenting a perceptually consistent metric for image error prediction, PieAPP. It is a state-of-the-art method to model human visual preference. Poster no. J4, Hall C-E, 12:30pm.
#cvpr
#CVPR2018
📢 A few weeks ago, I graduated with my Ph.D. from
@ucsantabarbara
.
Amidst the deep dive into perceptual metrics, visual saliency and media forensics, I gathered many life lessons and vital friendships. Very grateful for this journey and to everyone who helped me along the way!
[1/4] At
@BMVCconf
(Poster sess. 1), we will present noise-aware training, an effective strategy to train saliency predictors with limited gaze data, and a dynamic saliency dataset.
(w/ Iuri Frosio,
@0razio
, Pradeep Sen, Joohwan Kim,
@josefspjut
)
Teaser:
Please turn in your resume along with an intro to yourself to: eprashnani
@nvidia
.com
Note: this is an in-person, full-time opportunity, where the intern would be hosted in either the Santa Clara or the Boston NVIDIA office. The applicants need to have US work authorization.
I am honored to be giving a keynote at NeurIPS tomorrow!
I will present about the invisible workers who are driving our A.I. industry and discuss how we can co-design with them to create a more empowering future of work 👊🏾🔧
Join me!
👉🏾
#neurips2020
#AI
Presenting PieAPP (a perceptually consistent metric for image error prediction) at CVPR tomorrow! Find me at Hall C-E No. J4! See you there! :)
#cvpr2018
#cvpr
Clarifying an earlier call:
We are looking for a Ph.D. intern @ NVIDIA for research on human perception of 3D scenes. Human vision and/or graphics background with an interest in human vision. DM me if interested or have questions. Please spread the word.
Chief Scientist Bill Dally released an open-source design for a low-cost, easy-to-assemble mechanical ventilator. To move his project forward, he spent time gathering expert advice on the device’s capabilities and testing prototypes. Read more in detail on the design.
#COVID19
Nagging deadlines, fighting unproductive spells due to steep routine changes, many COVID-19 uncertainties...
And then: a moment of freedom from all these nags.
It came to me in the form of a bug in my code that consumed all of my attention. I am grateful.
#focusOnWhatYouCanDo
If interested, contact: eprashnani
@nvidia
.com.
Ideal applicant: a Ph.D. candidate with a proven track record of making insightful scientific contributions, and leading research publications at top-tier CV / ML venues. [2/2]
After my positive experience with the AI Grant, Pioneer is something to look forward to.
Glad to see such efforts to make financing, mentorship and peers accessible to all the lost geniuses across all areas of pursuit!
Way to go
@danielgross
and team!
@pioneerdotapp
Promising work at Nvidia Research by
@XueT_Li
@SifeiL
@shalinidemello
and team.
A self-supervised 3D reconstruction method using a single viewpoint of a scene...surpasses the need for manual labeling, multiple views, etc., & achieves results on par with supervised SOTA methods!
India's successful lunar mission (landing its Chandrayaan-3 spacecraft on the moon) is another one of many straws in the wind showing its ascendence. As previously shown in my health index for countries, which is used to derive my projections for countries' next 10-year growth
Maybe it’s deadline-induced sleep deprivation talking...but one of my favorite things to do is press ⏎ to start a shell script that will successfully launch several deep-learning experiments.
What a joy! 😃 ↩️
Reading Tesla’s writings and his biography by Bernard Carlson to mark his 77th death anniversary tomorrow...so humbled by his genius; his work and vision is truly inspiring!
Speaking of snowball effects...
Nikola Tesla’s impression of this phenomenon contributed to his mindset behind discovering the Tesla coil.
Following up on those little sparks could lead to profound discoveries: important reminder for the aspiring scientist in me. 😊
[2/4] Accuracy of saliency maps measured with gaze data varies with frame content (2 frames of a video shown below) & no. of available observers.
Hence, it is hard to guarantee high per-frame accuracy of measured per-frame saliency of dynamic videos, esp. with few observers.
[3/4] The traditional training of a saliency predictor is prone to noise overfitting since the discrepancy netween predicted & measured (noisy) saliency map is directly minimized.
With our method (NAT), we reduce overfitting to noisy maps by accounting for the noise in each map.
I followed Kobe Bryant as a source of inspiration for my own (sometimes seemingly insurmountable) struggles as a PhD student. What a legend - it seems unreal that life can be so very ephemeral.
#RIPKobeBryant
(Picture source: )
In about a week
@SIGGRAPH
, we will showcase an AI-Mediated 3D Video Conference system at Emerging Technologies, where you can talk to people in 3D using only a webcam + GenAI. It also features Live 3D Portrait and many more! Check it out
[4/4] We show the generalizability of NAT for visual saliency prediction with experiments across multiple datasets, dataset sizes, loss functions, & architectures.
NAT is particularly effective when limited gaze data is available.
It sucks not being able to visit your family
Just ask immigrant scientists who haven't seen their family in years because of visa issues & costs, and the very real possibility they would not be allowed back in the country
It's time to fix the US visa system
📢
#CVPR2023
Spontaneous Diffusion Meetup
Several folks have asked me to do a casual meetup for those interested in diffusion models
@CVPR
. Let's meet tomorrow (Wed) at 1 pm in West building (upper levels). Exact location will be posted below this tweet.
Please spread the word
@IndiainSpain
has been so instrumental in enabling my return trip to India from a recent visit to Spain. Really appreciate your efforts, kind sirs! Thank you for keeping the Indian Embassy open on a Sunday to help out those in need.
I think it is important for platforms of scientific discourse to be free of known sexual/racist connotations.
Please help make that happen by signing the petition!
#ProtestNIPS
thanks
@AnimaAnandkumar
@zacharylipton
2/ potential consequence of (B): too much backward referencing to the LA chapter might get a bit daunting if the chapter ends up being dense
Nothing beats the feeling of being accomplished than submitting a research paper... except maybe having a full night's sleep, but that's a story for a different time!