In Market.

My Contribution

I led Generate Presentation from research and proof of concept to production, and architected the system that turns a prompt and a source document into an audience-aware first draft in minutes — structured content generation, contextual imagery, charts, and rigorous evaluation. I cut time to first slide substantially without trading away content or visual quality, and mentored the engineers building it.

My Contribution

I lead the AI architecture behind PDF Spaces — multi-agent orchestration, prompt and workflow design, evaluation loops, and production reliability for collaborative document workspaces with intelligent document understanding.

My Contribution

I lead system design for contract understanding in Acrobat AI Assistant: prompt and workflow architecture, evaluation loops, and the production framework behind accurate clause analysis and extraction.

My Contribution

I shipped the on-device person detection and tracking powering Intelligent Zoom, Dynamic Framing and Virtual Director — automatic framing and active-speaker composition running in real time on the device. Hardware-in-the-loop optimization cut model memory by 16% and lifted performance by more than 20%.

My Contribution

The same on-device AI stack as the PanaCast 50 — Intelligent Zoom, Dynamic Framing and Virtual Director — tuned through quantization and hardware-in-the-loop optimization for the video bar system.

My Contribution

I built the on-device detection and segmentation behind Intelligent Zoom and Background Blur, taking both from research through optimization to a shipping product on constrained hardware.

My Contribution

I authored the core algorithm and shipped the first Intelligent Meeting Spaces release. Teams draw a meeting boundary, and only people inside it are framed — spatial awareness running on the device.

My Contribution

I contributed to medical-imaging AI for clinical decision support, supporting tools used by dental professionals to identify and review potential findings.

My Contribution

Built the on-device AI behind the Bose AR ecosystem and delivered Bose's first mobile on-device neural network, shipped in the BoseAR library. Designed privacy-first ML pipelines and worked with hardware platform teams to run inference on battery-powered devices.

My Contribution

Trained a model that recognizes ten gestures from head motion alone — no camera, no microphone, just the 9-axis sensor stack in Bose Frames. Real-time gesture detection on battery-powered hardware.