• Architected multi-modal AI agents by fine-tuning LLMs for real-time applications in education and automated retail (QSR) environments.

• Developed custom agentic visualization tools to dynamically generate diagrams and imagery, enhancing conceptual learning for users.

• Optimized low-latency Human-AI interaction by deploying real-time STT/TTS pipelines across Android, iOS, and Web platforms.

• Bridging the Sim-to-Real Gap: Architected end-to-end PBR synthetic data pipelines for human-centric ML models, reducing training error by matching the statistical distribution of real-world sensor noise and lighting.

• Sensor & Lens Fidelity: Engineered a proprietary library for extracting ray-accurate sensor models from CodeV/Zemax, enabling the simulation of cameras through pancake lens systems.

• Neural Ground Truth Generation: Developed hybrid data workflows using 3DGS and NeRFs to reconstruct high-fidelity "digital twins," providing pixel-perfect ground truth for depth, occlusion, and semantic segmentation.

• Scalable Rendering Architecture: Optimized large-scale workflows (OpenCue, custom renderfarms) to accelerate synthetic dataset throughput, ensuring consistent multimodal metadata across RGB, Depth, and IR sensors.

• Research Collaboration: Partnered with research teams to support peer-reviewed submissions in human motion synthesis and generative 3D modeling, with accepted papers at ECCV, WACV, and FG.

• Spatial Computing & Real-time Rendering: Contributed to asset/shader loading and animation components (OpenXR, Vulkan) for the Meta Spatial SDK, and implemented experimental algorithm prototypes for high-fidelity VR headset passthrough.

• Led development of "Torch," a renderer-agnostic node-graph lighting tool used across the studio’s feature film slate.

• Managed full-lifecycle feature delivery as Scrum Lead, coordinating cross-functional scheduling and UI/UX enhancements for the lighting department.

• Refactored core node-graph logic to improve UI responsiveness and usability, significantly reducing artist iteration time on complex lighting setups.

• Developed real-time GPU-based lighting tools, utilizing deep pixel buffer caches to significantly reduce render-loop latency for real-time visual effects.

• Optimized production performance by migrating legacy Renderman shaders and C++ shade ops to modern GLSL, achieving substantial gains in interactive preview speeds.

• Collaborated as a core member of the team to design and develop Katana, the industry-standard 3D Lighting & Look-Dev tool.

• Engineered procedural scene resolution logic, enabling artists to override complex 3D scene graphs at scale—a workflow now used globally in VFX.

• Architected foundational APIs for image and scene-graph processing, optimizing the pipeline for high-performance 3D asset evaluation.