Ajay U. Mandlekar (original) (raw)
Contact
Email: amandlekar@nvidia.com
Twitter: @AjayMandlekar
Google Scholar: Ajay Mandlekar
Github: amandlek
CV: link
Recent News
- [April 2025] We have released the DexMimicGen simulation environments and datasets and the AHA codebase.
- [April 2025] 3 papers presented at ICLR 2025: MimicLabs (a study in collecting and using large-scale manipulation datasets), AHA (novel failure data generation and failure reasoning framework for robotic manipulation), and LAPA (unsupervised latent action based pretraining for robotic manipulation).
- [March 2025] Check out Sim-and-Real Co-Training, a simple recipe that utilizes synthetic data generation tools (such as MimicGen) for efficient real-world robotic manipulation. This recipe was used to train the GR00T-N1 humanoid foundation model as well.
About Me
I am a research scientist at NVIDIA AI. I obtained my PhD degree from Stanford University, co-advised by Silvio Savarese and Fei-Fei Li. My research focuses on developing systems and algorithms to enable robots to learn useful manipulation tasks.
Education
- Ph.D., Electrical Engineering, Stanford University co-advised by Silvio Savarese and Fei-Fei Li, September 2016 - January 2022
- M.S., Computer Science, Stanford University, June 2018
- B.S., Electrical Engineering, California Institute of Technology, June 2016
- B.S., Computer Science, California Institute of Technology, June 2016
- Monta Vista High School, Class of 2012