cv

M.Sc. in Neural Information Processing

• Grade - "Excellent" | 3.76/4 (US Scale) | 1.24/1 (German Scale).

B.Tech. in Electrical Engineering

• Specialization in Cognitive and Intelligent Systems.
• Grade - 8.6/10.

Researcher

• Supervisors - Prof. Kai-Wei Chang and Prof. Aditya Grover.
• Objective - Defending multimodal contrastive learning models (e.g., CLIP) against backdoor attacks.
• Demonstrated the effectiveness of combining multi-modal contrastive loss (e.g., CLIP) and uni-modal self-supervision loss (e.g., SimCLR) for fine-tuning on limited clean data, countering attacks while maintaining high downstream performance.

Researcher

• Supervisor - Dr. Wieland Brendel.
• Objective - Develop a regularization approach for training neural networks that ensures each neuron aligns with a distinct semantic concept, thereby enhancing the mechanistic interpretability of deep neural networks.
• We associate each neuron with a specific concept represented by a point in the CLIP embedding space. Then, we train the neural network to position highly activating images close to the concept descriptor within the CLIP embedding space.

Researcher

• Supervisor - Prof. Falk Lieder.
• Objective - Create a theoretical framework for setting subgoals that can assist individuals in achieving their objectives.
• We use a computational model of human behavior to estimate how much a subgoal benefits individuals. Then, we use optimization methods to identify subgoals that optimize performance. Our subgoals improve performance in user studies.

Visiting Researcher

• Supervisor - Prof. Tao Gao.
• Objective - Learn policies for multi-agent adversarial games in MuJoCo using Deep Reinforcement Learning.
• Implemented AlphaZero and benchmarked its performance against model-free, tree-search algorithms, and heuristic policies.

• Best Paper Award ($1,000) as a co-first author for CleanCLIP at the RTML Workshop at ICLR, 2023.
• Diversity and Inclusion Award ($1,000) - Among 10 recipients (worldwide) awarded by Cognitive Science Society.

• Bounded Rationality Winter School - Among 40 selected (worldwide) for winter school organized by MPI Berlin.

• Prof. R.K. Mittal Award (INR 10,000) - Given to 2 freshmen (out of 850) at IIT Delhi for academic performance.
• IIT Delhi Merit Award - Conferred for securing a position among the top 7% students of the batch at IIT Delhi.

• IIT-JEE - Ranked amongst the top 0.01% applicants out of 1.5 million candidates in IIT-JEE Advanced.