Industry & research, in parallel

Researching and developing a real-time Talking Head system — a conversational AI with a live, responsive face that lets users hold natural, face-to-face voice conversations with AI.

• Researching and building a real-time Talking Head system for live, face-to-face AI conversations
• Designing low-latency streaming pipelines for natural, real-time speech-to-speech interaction
• Integrating conversational LLMs with synchronized facial animation for a responsive AI avatar
• Prototyping the real-time audio/video streaming architecture end to end, from research to implementation
• Iterating on model and system performance to keep conversations fluid under real-time constraints

Conducting applied AI/ML research spanning efficient LLM deployment and real-time video understanding — co-authored Z-Pruner (IEEE AICCSA 2025) and developed TimeClipAI, a real-time action localization framework.

• Co-authoring Z-Pruner, a post-training LLM pruning method published at IEEE AICCSA 2025
• Developing TimeClipAI, a real-time temporal action localization framework with state-of-the-art benchmark results
• Implementing and benchmarking deep learning models in PyTorch
• Reading, reproducing, and extending state-of-the-art research papers
• Collaborating with faculty members and co-authors across research labs