Bhavika Devnani

I'm a first-year PhD student at the Hoffman AI Research Lab at Georgia Tech, advised by Dr. Judy Hoffman. I've been incredibly lucky to work on different facets of machine learning problems at Apple AI/ML Research, Quora, and LinkedIn.

I enjoy thinking about Multimodal Representation Learning and Efficiency.

Email | CV | Google Scholar | GitHub

Bhavika Devnani
News
[May 2025] Interning with Apple AI/ML for the summer.
[Oct 2024] ELSA accepted at NeurIPS 2024!
[Aug 2024] Started a PhD at Georgia Tech with Dr Judy Hoffman.
[Dec 2022] Started working at Apple AI/ML Research - led by Dr Samy Bengio.
[Oct 2022] ZSON accepted at NeurIPS 2022!
[Nov 2022] BiSA wins best paper at NeurIPS 2022 - Vision Transformers Workshop!
[Oct 2022] BiSA accepted at NeurIPS 2022 - Vision Transformers Workshop!
Research
ELSA Paper
ELSA: Learning Spatially-Aware Language and Audio Embeddings

Bhavika Devnani, Skyler Seto, Zakaria Aldeneh, Alessandro Toso, Yelena Menyaylenko, Barry-John Theobald, Jonathan Sheaffer, Miguel Sarabia

Accepted at NeurIPS 2024

Generated dataset and trained a model to align 3D spatial audio with open vocabulary captions.

ZSON Paper
ZSON: Zero-Shot Object-Goal Navigation using Multimodal Goal Embeddings

Arjun Majumdar*, Gunjan Aggarwal*, Bhavika Devnani, Judy Hoffman, Dhruv Batra

Accepted at NeurIPS 2022

CLIP enables Zero-Shot Object-Goal Navigation by learning multimodal goal embeddings.

BiSA Paper
Bi-Directional Self-Attention for Vision Transformers

George Stoica, Taylor Hearn, Bhavika Devnani, Judy Hoffman

Accepted at NeurIPS 2022, Vision Transformers Workshop - Best Paper

Refined sources based on surrounding context by inverting self-attention.

"The struggle itself towards the heights is enough to fill a man's heart. One must imagine Sisyphus happy." - Albert Camus