Conducted research on few-step generative models, diffusion models, representation learning and neural compression in the Signals, Information and Algorithms group. Defended my thesis title “Score Estimation for Generative Modeling”. Advised by Professor Gregory Wornell.

Completed my thesis titled “Image Compression using Sum-Product Networks” under supervision of Professor Gregory Wornell.

Graduated with Highest Honors and the Bronze Tablet. Completed my undergraduate thesis titled “LAP-based Motion Compensated Frame Interpolation” under supervision of Professor Pierre Moulin.

Working on codec training, few-step generative modeling, reward modeling and LLM post-training for multimodal LLMs.

Worked with Dr. Nikos Vlassis on one-step image generative modeling from scratch.

Interned on the Neural Compression team, where I worked with Dr. Fabian Mentzer on neural video compression using vision transformers.

Interned on the AI Speech team, where I worked with Dr. Vimal Manohar and Dr. Qing He on self-supervised representation learning for singing voice conversion using HiFi-GAN vocoders.

Interned on the AI Speech team, where I worked with Dr. Qing He on variable bitrate neural speech codecs using hierarchical VQ-VAEs and WaveRNNs with configurable computation complexity.

Interned with Dr. Jonathan Le Roux and Professor Pierre Moulin from UIUC, where I worked on developing novel algorithms for detecting adversarial attacks on speech recognition systems.

Worked on deep learning graph optimization and profiling using TensorRT on edge devices.