Experience

Research Scientist · 3D Foundation Models

• Work on training, data, and evaluation for Meshy's 3D generation models, including Meshy-7.

Research Engineer · Data for Embodied AI

• Developed multiview human motion-capture pipelines and trained custom models for pose estimation and hand tracking, producing motion data for embodied-AI and humanoid research.
• Shipped Embody3D: built QA tooling and automated quality filters, coordinated 1.6M video annotations, and curated 2,000 hours of motion data across body, hands, audio, and text.
• Cut keypoint tracking error by 75% with a self-supervised data-generation pipeline.

Research Engineer Intern · 3D Foundation Model Training

• Built a pipeline converting synthetic mesh assets into relightable Gaussian splats for 3D foundation model training.
• Wrote custom CUDA kernels replacing spherical harmonics with neural features; scaled to 32 nodes, taking per-asset processing from 6 minutes to 30 seconds.

Research Scientist Intern · Explainable Recommendation

• Counterfactual explanations for visual recommenders: minimal image-embedding perturbations, grounded in language with CLIP.
• Published at WWW 2024 (paper); USPTO patent filed.

Machine Learning Intern · Document Understanding

• End-to-end multilingual invoice understanding: table extraction with object detection, field extraction with LayoutLMv2.

MS in Robotics · GPA 4.11/4.0 · Advisor: Prof. Ioannis Gkioulekas

• Coursework: Deep Learning Systems, Learning for 3D Vision, Robot Learning, Physics-based Rendering.

BTech in Electrical Engineering · Minor in Machine Learning · CGPA 9.24/10

• Coursework: Modern Computer Vision, Machine Learning, Computational Photography.