About Me

Hi! I am Vishwa Shah, I am a graduate student at LTI, CMU in the MIIS program. Recently, I completed my AI Residency  at Meta California, where I first worked on low-resource Multimodal detection of integrity violations in VR and later on inducing personality traits in LLaMA using reward-modeling.

Currently my research focuses on reducing Hallucinations and improving Faithfulness in LLM generations.  I am working with Prof. Maarten Sap  on mitigating intrinsic hallucination by reinforcement from fine-grained feedback. With Prof. Graham Neubig , I am working on analyzing robustness of LLMs to position of information in long-context.

I am also interested in research around techniques to improve transferability and adaptability for low-resource multilingual and multimodal settings.

I graduated with a B.E. (Hons.) Computer Science from BITS Pilani, K K Birla Goa Campus, India  in 2022. I have had the opportunity to work as a research intern at MIDAS in the domain of Multimodal AI and at APPCAIR on Neural Algorithmic reasoning. As an Applied Scientist intern at Amazon India, I got to explore video understanding tasks. I have had the pleasure of working as a SWE intern at Google India and Microsoft India (R&D) providing me a chance to explore the Software Engineering realm.

Additionally, I have volunteered for WiMLDS, and am a member at SAiDL.