Sanjana Srivastava
Stanford, California, United States
2K followers
500+ connections
About
Graduating Stanford AI PhD student interested in evaluation and RL-based adaptation of…
Activity
-
Shout out to Ashu Garg, the person I connected with the most on LinkedIn this year. #YearinReview 😂😂😂
Shout out to Ashu Garg, the person I connected with the most on LinkedIn this year. #YearinReview 😂😂😂
Liked by Sanjana Srivastava
-
Was surprised to see an article about caregiving costs on my Yahoo Finance homepage today. (Yeah, I know—who still uses Yahoo?) The…
Was surprised to see an article about caregiving costs on my Yahoo Finance homepage today. (Yeah, I know—who still uses Yahoo?) The…
Liked by Sanjana Srivastava
-
We hosted a fun NeurIPS event in San Diego with Inception, Karan Singhal, and Turing. Thank you to everyone that came and we are excited to be back…
We hosted a fun NeurIPS event in San Diego with Inception, Karan Singhal, and Turing. Thank you to everyone that came and we are excited to be back…
Liked by Sanjana Srivastava
Experience
Education
Volunteer Experience
-
Teacher
LearningWorks at Blake: A Breakthrough Collaborative Program
Publications
-
ROSETTA: Constructing Code-Based Reward from Unconstrained Language Preference
Workshop on Continual Robot Learning from Humans @ Robotics: Science and Systems, Workshop on Human-in-the-Loop Robot Learning
Developed method to generate code-based RL reward from individual unseen language preferences. Works one-shot in difficult settings like ManiSkill 3 continuous control. Designed thorough metric suite. 86% human satisfaction, 87% task success rate on dataset of 116 unique preferences.
-
BEHAVIOR-1K: A Benchmark for Embodied AI with 1,000 Everyday Activities and Realistic Simulation
Conference on Robot Learning
See publicationBuilt the largest embodied AI benchmarks of layperson-defined tasks in various physical modalities. Designed data-driven human annotation pipeline, implemented logic solver and analytical form to get automatic binary evaluation and non-sparse reward. Conducted 2,000-person task preference survey and scaled to 1,000 tasks in OmniGibson simulator. Extensive code/engineering/design collaboration.
-
BEHAVIOR: Benchmark for Everyday Household Activities in Virtual, Interactive, and Ecological Environments
Conference on Robot Learning, 2021
-
Identifying Learning Rules from Neural Network Observables
Neural Information Processing Systems
-
A Machine Learning-Based Analysis of Game Data for Attention Deficit Hyperactivity Disorder Assessment
Games for Health Journal
A Machine Learning-Based Analysis of Game Data for Attention Deficit Hyperactivity Disorder Assessment. Monika D. Heller, Kurt Roots, Sanjana Srivastava, Jennifer Schumann, Jaideep Srivastava, and T. Sigi Hale. Games for Health Journal. October 2013, Vol. 2, No. 5: 291-298
Other authorsSee publication
Patents
-
Predictive Executive Functioning Models Using Interactive Tangible-Graphical Interface Devices
Filed US US20140057244 A1
A method of diagnosing attention deficit hyperactivity disorder (ADHD). The method employs a tangible graphical interactive game wherein the interactive game employs a plurality of tangible graphical cubes. The game induces stimuli, measures responses and accumulates the responses using a predefined set of variables into a predefined set of metrics, wherein the variables are determined using an interactive machine learning feedback algorithm.
Other inventorsSee patent
Projects
-
ROSETTA: code reward from unconstrained language
-
Developed method to generate code-based RL reward from individual unseen language preferences. Works one-shot in difficult settings like ManiSkill 3 continuous control. Designed thorough metric suite. 86% human satisfaction, 87% task success rate on dataset of 116 unique preferences.
-
BEHAVIOR-100 and BEHAVIOR-1K benchmarks
-
Built the largest embodied AI benchmarks of layperson-defined tasks in various physical modalities. Designed data-driven human annotation pipeline, implemented logic solver and analytical form to get automatic binary evaluation and non-sparse reward. Conducted 2,000-person task preference survey and scaled to 1,000 tasks in OmniGibson simulator. Extensive code/engineering/design collaboration.
-
Offline RL with binary preference
-
Implemented offline actor-critic and offline MuZero advantage objectives and used years of preference-annotated demos from DM Playhouse to train advantage directly. Designed tasks with long-horizon dependencies to evaluate offline RL compared to online setter-replay.
Honors & Awards
-
ROSETTA Best Paper Award
Workshop on Continual Robot Learning from Humans @ Robotics: Science and Systems
-
BEHAVIOR-1K Best Paper Nomination
Conference on Robot Learning (CoRL)
-
NSF Graduate Research Fellow
National Science Foundation
-
English Achievement Award
-
Award for best student in the subject of English across school.
-
National Merit Semifinalist
-
-
National AP Scholar
-
-
North Star Journalist Award
-
State-wide award for excellence in journalism; accepted for work in school newspaper.
-
Siemens Competition Semifinalist
Siemens Foundation
Placed as a semifinalist in the annual Siemens Competition in Math, Science, and Technology.
-
Startup Awards for CogCubed
Minnesota High Tech Association; Minnesota Cup
CogCubed, Inc. received recognitions for presentations based on the technology I developed (Tekne Award finalists in Hi-Tech and Healthcare startup categories; Minnesota Cup High-Tech division runner up).
-
American Mathematics Competition 10 School Top Scorer
-
-
American Invitational Mathematics Examination Qualifier (2012)
-
-
Grace Hopper Poster Presenter and Scholar
Anita Borg Institute
Poster accepted at the annual Grace Hopper Celebration, one of the foremost conventions for women in computer science. Separate scholarship received to attend conference from the Anita Borg Institute, the host of GHC.
Languages
-
English
Native or bilingual proficiency
-
Hindi
Native or bilingual proficiency
-
Mandarin
Elementary proficiency
More activity by Sanjana
-
We've been running RadixArk for a few months, started by many core developers in SGLang and its extended ecosystem (slime, AReaL). I left xAI in…
We've been running RadixArk for a few months, started by many core developers in SGLang and its extended ecosystem (slime, AReaL). I left xAI in…
Liked by Sanjana Srivastava
-
Best part about NeurIPS for me was to see our work featured by Yejin Choi at the Keynote.
Best part about NeurIPS for me was to see our work featured by Yejin Choi at the Keynote.
Liked by Sanjana Srivastava
-
I will join Northwestern University Computer Science as an Assistant Professor in Fall 2026. I am actively recruiting PhD students this cycle and…
I will join Northwestern University Computer Science as an Assistant Professor in Fall 2026. I am actively recruiting PhD students this cycle and…
Liked by Sanjana Srivastava
-
UC Berkeley and UCSF researchers have pushed the frontier of AI for radiology with the release of Pillar (Pillar-0), an open-source AI model that…
UC Berkeley and UCSF researchers have pushed the frontier of AI for radiology with the release of Pillar (Pillar-0), an open-source AI model that…
Liked by Sanjana Srivastava
-
I’m excited to share two key milestones for Parallax Worlds today: a new funding round led by Midas list investors, and an industrial-grade…
I’m excited to share two key milestones for Parallax Worlds today: a new funding round led by Midas list investors, and an industrial-grade…
Liked by Sanjana Srivastava
-
🚀 Excited to share the release of our recent work on 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗥𝘂𝗯𝗿𝗶𝗰𝘀 – a benchmark for evaluating Deep Research (DR) agents across…
🚀 Excited to share the release of our recent work on 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗥𝘂𝗯𝗿𝗶𝗰𝘀 – a benchmark for evaluating Deep Research (DR) agents across…
Liked by Sanjana Srivastava
-
Had a fantastic time at the National Association of Social Workers Ohio Chapter Annual Conference last week. We met 500 social workers from across…
Had a fantastic time at the National Association of Social Workers Ohio Chapter Annual Conference last week. We met 500 social workers from across…
Liked by Sanjana Srivastava
-
yesterday, Monica Xie and I hosted our 2nd roundtable with folks from NVIDIA, Google DeepMind, Dyna Robotics, Stanford University, and more to talk…
yesterday, Monica Xie and I hosted our 2nd roundtable with folks from NVIDIA, Google DeepMind, Dyna Robotics, Stanford University, and more to talk…
Liked by Sanjana Srivastava
-
“Creativity is intelligence having fun.” Unleash your creativity and imagination with Marble - our 3D world generation model, now available to…
“Creativity is intelligence having fun.” Unleash your creativity and imagination with Marble - our 3D world generation model, now available to…
Liked by Sanjana Srivastava
-
📢 Calling all prospective PhD students interested in wearable robotics and movement biomechanics! 🎉 I’m excited to share that I will be joining…
📢 Calling all prospective PhD students interested in wearable robotics and movement biomechanics! 🎉 I’m excited to share that I will be joining…
Liked by Sanjana Srivastava
Other similar profiles
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Sanjana Srivastava
144 others named Sanjana Srivastava are on LinkedIn
See others named Sanjana Srivastava