• Lead cross-functional team to research, build, and drive adoption of deep learning vision models for single cell detection. Achieved 33% higher candidate counts and 0.7 percentage points lower doublet rates while eliminating user-facing analysis wait time. Use Python, PyTorch, and OpenCV for model development.

• Lead cross-functional team to research and build generative AI LLM agents and ML models to accelerate development and analysis of liquid handler chemistry protocol automation.

• Lead effort to research, build, and train AI optimization models using reinforcement learning. Utilize Python, PyTorch, Stable Baselines3 (SB3), and Gymnasium for model development.

• Architected and developed custom AI inference engine for serving fast, real-time model inference across all Takara instruments, with 10x faster I/O speed and at-rest encryption for proprietary IP on deployed hardware. Used Python, gRPC, ONNX, and OpenVINO.

• Develop computer vision algorithms for automated high-precision instrument calibration (auto-focus, alignment) and checks (verifying proper loading and reading IDs of consumables) using OpenCV and RapidOCR.

• Built and manage on-premises AI compute cluster using K3s, dstack, and MLflow to enable internal R&D model training and inference as well as experiment tracking.

• Responsible for establishing AI standards and delivering seminars highlighting and advocating for AI initiatives at Takara. Lead all instrumentation-related AI R&D efforts.

• Architected and developed control software for qPCR liquid handler instrument, utilizing C#, WPF, MVVM, Redux-like state management, and SQLite. Serve as sole technical product owner, taking user feedback, supporting in-house experiments, orchestrating cross-functional collaboration, and communicating with field service team to support in-field instruments.

• Designed modern UI/UX style system for use across Takara’s qPCR and single cell instrumentation using Figma.