Hafeez Khan

Welcome! I am a PhD student specializing in how foundation models adapt under distribution shift, addressing challenges in continual learning and few-shot learning. My advisor is Siddhartha Bhattacharyya, and I am funded by NASA, where I am working on lane detection for autonomous driving and taxiing on airports.
Previously, I completed my undergraduate studies at BITS Pilani Dubai, where I finished my thesis under the guidance of Raja Muthalaghu and received the Outstanding Undergraduate Thesis Award. Alongside my PhD, I have worked closely with mentors from Microsoft Research, NASA Langley, UTHealth Houston, Collins Aerospace. Through these collaborations, I have gained extensive experience especially in self-supervised models, diffusion models, and multimodal large-language models.
Reach out by email if our interests align and you wish to collaborate!
Also, I’m actively seeking an internship for Summer 2026—because models aren’t the only ones that need continual learning!!
news
Aug 21, 2025 | Recognized for outstanding academic achievement and invited to join Upsilon Pi Epsilon (UPE), the international honor society for computing sciences! |
---|---|
Jul 13, 2025 | Test-time Prompt Refinement for Text-to-Image Models accepted at ICCV 2025 (MARS2 Workshop, Oral Presentation)! |
Jul 11, 2025 | Adapt But Don’t Forget accepted at ICCV 2025 (2COOOL Workshop, Poster Presentation)! |
selected publications
- Test-time Prompt Refinement for Text-to-Image ModelsIn Workshop on Multimodal Reasoning and Slow Thinking in Large Model Era at ICCV, 2025
- NORA: A Nephrology-Oriented Representation Learning Approach Towards Chronic Kidney Disease ClassificationIn Under Submission at International Conference in Machine Learning and Applications (ICMLA), 2025
- Adapt, But Don’t Forget: Modular Fine-Tuning and Contrastive Routing for Lane Detection Under Distribution ShiftIn Workshop on Out Of Label Hazard Detection in Autonomous Driving at ICCV, 2025
- Modular Test-time Input-Space Refinement for Few-Shot SegmentationIn Under submission in Neural Information Processing Systems (NeurIPS), 2025
- ALINA: Advanced Line Identification and Notation AlgorithmIn Workshop on Vision Datasets Understanding at CVPR, 2024