About me

I’m a Ph.D. student in Computer Science at Georgia Tech, where I’m fortunate to be advised by Prof. Judy Hoffman. My research focuses on developing multimodal vision-language models that integrate spatial, semantic, and temporal reasoning with minimal supervision.

Recent work includes:

1. A 7B open-source VLM for open-vocabulary 3D scene graph generation, under review at NeurIPS 2025.

2. SkyScenes, a synthetic aerial dataset for improving real-world segmentation, accepted at ECCV 2024.

3. Generalist Multimodal LLM, where I designed a jointly-trained vision-audio model that outperforms larger generalist systems by reducing cross-modal interference.

I bring prior experience in domain generalization, zero-shot learning, and synthetic-to-real adaptation, focusing on making models robust to diversity, correlation, and semantic shifts in real-world environments. My goal is to build generalizable systems that require minimal labeled data yet remain reliable under distribution shifts.

I also review for top conferences (NeurIPS, CVPR, ECCV), and have published across both vision and language communities.

🌟 I’ll be attending CVPR 2025 — if you’re around and working in similar areas, I’d love to connect!

💼 I’m currently looking for research internships for Summer 2026 — feel free to reach out if you’re hiring!

Recent Updates

[ 🌟: Important | 💡: Research Paper | 📆: Miscellaneous ]

• 🌟 Attending CVPR 2025 in Nashville!

• 🌟 Attending ECCV 2024 at Milan, Italy! Georgia Tech published an article about our work!

• 💡 Jul 1, 2024: My first first-author paper – SkyScenes: A Synthetic Dataset for Aerial Scene Understanding is accepted at ECCV 2024!

• 🌟 Apr 1, 2024: Joining Georgia Tech for Ph.D. CS under Prof. Judy Hoffman in Fall 2024.

• 📆 Mar 12, 2024: Serving as a reviewer for ECCV 2024.

• 💡 Oct 24, 2023: My first main-conference paper - LatentDR: Improving Model Generalization Through Sample-Aware Latent Degradation and Restoration is accepted at WACV 2024!

• 🌟 Attending NeurIPS 2022 at New Orleans, LA, USA. My first in-person conference!

• 🌟 Apr 4, 2022: Admitted to the MS CS program of Georgia Tech for Fall 2022!

Publications

2024

• ECCV 2024 [First first-author paper!]
  • SkyScenes: A Synthetic Dataset for Aerial Scene Understanding
  • Sahil Khose*, Anisha Pal*, Aayushi Agarwal*, Deepanshi*, Judy Hoffman, Prithvijit Chattopadhyay
  • Website | HF🤗 | Paper | CoC Article | GitHub
• WACV 2024 [First main-conference paper!]
  • LatentDR: Improving Model Generalization Through Sample-Aware Latent Degradation and Restoration
  • Ran Liu, Sahil Khose, Jingyun Xiao, Lakshmi Sathidevi, Keerthan Ramnath, Zsolt Kira, Eva L. Dyer

2022

• NeurIPS 2022 [Visted my first in-person conference in New Orleans!]
  • [Poster] Continual VQA for Disaster Response Systems at CCAI
• ICML 2022
  • [Best Paper Award 🌟] An Efficient Modern Baseline for FloodNet VQA at New In ML
• ACL 2022
  • [Oral] Transformer based ensemble for emotion detection at WASSA

2021

• NeurIPS 2021
  1. [Spotlight Paper 🌟] Semi-Supervised Classification and Segmentation on High Resolution Aerial Images at CCAI
  2. XCI-Sketch: Extraction of Color Information from Images for Generation of Colored Outlines and Sketches at 1. [Oral] New in ML 2. [Paper] CtrlGen 3. [Paper] ML4CD 4. [Poster] DGM
  3. [Poster] A Studious Approach to Semi-Supervised Learning at ICBINB
• NAACL 2021
  • [Oral] BERT based Transformers lead the way in Extraction of Health Information from Social Media at SMM4H