Nikhil Behari

What can you see if you keep slowing down time? @veritasium explores the answers to that question with Media Lab PhD student @NikhilBehari (@cameraculture), Dr. Jim Bales (@MITEdgerton), and other experts. Watch: youtube.com/watch?v=P-4pbF…

@veritasium @raskarmit @medialab @anagh_malik 🎥 Watch the full video here: youtu.be/P-4pbFcERnk?t=… 🔬 Check out my work in this area: nikhilbehari.github.io

Huge thanks to Gregor and the @veritasium team for including me in their video, and to @raskarmit, Corey Ippolito, and the @medialab for their continued support of my research in this exciting area. Also shoutout @anagh_malik—check out his light-in-flight visuals in the video!

It was an incredible experience to be featured on my favorite science channel, @veritasium, to demonstrate how we can visualize light in ultra-slow motion, i.e. "light in flight."⚡️

@veritasium @DaveLindell @kyroskutulakos @GordonWetzstein @Po_lhr @NoahJuravsky Also cheers to Gregor Čavlović and the @veritasium team (see their tweet below) for making the video—and @NikhilBehari for the great explanation of light-in-flight imaging! x.com/veritasium/sta…

What if you keep slowing down?

Nikhil Behari, a graduate student in the MIT Media Lab Camera Culture group, designs new AI systems that can see more from pictures and turn a photo’s shadows and reflections into insights about the surrounding world.

Thanks @CSProfKGD for the shout out! It was great to demo looking around corners in real-time with just a $30 LiDAR! If you missed it, stop by our poster 77 today afternoon to learn more! #CVPR2025

Special thanks to my collaborators Aaron Young, @sidartsoma, @tklinghoffer, @adaveiitm and @raskarmit. We'll be presenting in Nashville this week -- feel free to reach out or stop by our poster!

Excited to present our #CVPR2025 ✨Highlight Paper "Blurred LiDAR for Sharper 3D." We introduce a method for 3D Gaussian surfel reconstruction that combines RGB images with diffuse (blurred) LiDARs. Surprisingly, we find that these blurred LiDARs, often used as simple proximity sensors, can improve reconstruction quality over conventional mobile LiDARs for robust handheld scanning in challenging scenes. 📍 Come by Sunday, June 15 · ExHall D · Poster #77 🔗 Website: nikhilbehari.github.io/bls3d-web/

@adaveiitm @stonybrooku Congrats Akshat!

📢 I'm happy to share that I'll be joining @stonybrooku as an Assistant Professor in CS this Fall. I will start the Photon Intelligence Lab, developing physics-aware AI that sees what we cannot – by bridging optics, graphics, and vision. Come join us! DM or meet me @ CVPR 🧵👇

Key takeaways? 🔧Adaptable: Dynamically adjusts to shifting public health priorities via language-guided inputs. 🌎Grounded: Leverages multi-agent RMAB simulations to anchor LLM-generated proposals. 🔁Self-improving: Iteratively enhances reward alignment with policy goals through LLM-driven self-reflection. Read more: arxiv.org/pdf/2402.14807 and find our poster at #NeurIPS2024!

How does it work? DLM uses human-language inputs to dynamically shape RMAB policies. From human input prompts, DLM generates reward functions aligned with policy goals, simulates policies learned under each proposed reward in multi-agent simulation environments, and then iteratively refines proposals by analyzing simulation outcomes.

🚨 New at #NeurIPS2024: Decision Language Model (DLM). We enable flexible, language-driven control of resource allocation in public health by leveraging principled RMAB decision-making with LLM-powered reward function design. Check out our poster next Wednesday at 11am PST in East Exhibit Hall A-C 2505! Check out our project page: nikhilbehari.github.io/dlm-web/ @MilindTambe_AI, Edwin Zhang, @yunfan_zhao, Aparna Taneja, @draj_uplink

Excited and honored for this opportunity to work in collaboration with @NASA at the MIT @medialab!

📢 Looking to recruit 3 PhD students to join my Algorithmic Decision Making and Society group @HPI, Germany. If you are interested in computational social choice, fairness & transparency in AI, or AI for public health, please consider applying by July 12: niclas-boehmer.com/jobs

Want to learn how to turn objects into cameras? Come to our #CVPR2023 poster THU-PM-016 presented by @NikhilBehari and @tklinghoffer CVPR link: cvpr2023.thecvf.com/virtual/2023/p… Project page: ktiwary2.github.io/objectsascam/