I build thoughtful AI systems and playful interfaces. I’m an NYU M.S. in Computer Science, after a double-major in Electrical Engineering and Finance.

I designed end-to-end ML infrastructure on AWS SageMaker, packaged PyTorch to ONNX with dynamic quantization, wrapped to Core ML for sub-second, on-device inference, containerized with Docker, and deployed to Kubernetes. I also orchestrated diffusion workflows—LoRA/DreamBooth, ControlNet—across data collection and generation pipelines.

I built a drug–excipient compatibility expert system achieving 97% accuracy, trained masked autoencoders for X-ray representation, created CLIP-style text–image embeddings, and developed a sequence-to-sequence approach that fused marginal structural models with representation learning—yielding a 48.1% lift.

I'm currently focused on Agentic AI applications and video representation learning—JEPA, masked video diffusion, and CLIP-style contrastive training— aiming for generalizable features that rival supervised baselines on targeted subtasks.