Selected Projects

Innovations in Medical AI: Transforming Cancer Detection at Google

In my tenure at Google Health, I played a pivotal role in a few groundbreaking projects including enhancing AI models for lung cancer detection and advancing AI in mammography for breast cancer screening.

My contributions as a technical lead included both leadership and technical aspects including the meticulous evaluation of AI models against human subjects in diverse real-world scenarios, improving the model performance and ensuring robust model generalizability across different patient demographics.

The lung cancer detection project, using 3D volumetric modeling of CT scans, achieved remarkable accuracy aiding radiologists for both lung cancer screening and Incidental Pulmonary Nodules (IPN). Google Health partnered with Aidence for further development and market introduction. For more details, you can explore lung cancer prediction blog post and the lung cancer work with Aidence here.

In the breast cancer initiative, we focused on integrating AI into mammography workflows, significantly improving the detection accuracy and consistency. Google health partnered with iCAD to incorporate the technology into their ProFound Breast Health Suite. For more information on advancing breast cancer detection with AI please visit here.

Our efforts in both areas have culminated in several papers being accepted for publication and others under review, demonstrating the impactful nature of our research.

Assistive AI in Lung Cancer Screening: A Retrospective Multinational Study in the United States and Japan

Harmonizing Search and AI: Enhancing Amazon Music with Machine Learning

At Amazon Music, I played a key role in enhancing the user experience by developing a machine learning-based ranking algorithm for the search functionality. By improving search accuracy and relevance, the ranker makes Alexa interactions smoother and more intuitive, allowing users to find their desired music with minimal effort.

As a Machine Learning Engineer, I collaborated closely with the applied science team to integrate our ML model into the live production environment. This integration required extensive engineering work, including setting up data streams, deploying the ML model in production, conducting A/B testing, launching the system, and continuously monitoring performance through dashboards.

Highlighting my innovation in this field, I also led a team to a notable victory in Amazon Music's ML hackathon.

Ultrasound Innovations in Prostate Cancer: My Research Journey at Stanford

At Stanford University as a Postdoctoral Researcher, I contributed to prostate cancer treatment research, particularly in using ultrasound for treatment monitoring.

Real-time tracking in prostate cancer treatment is important because it ensures radiation precisely targets the tumor, accounting for organ movement during therapy. This accuracy reduces radiation exposure to healthy tissues, minimizing side effects and enhancing treatment effectiveness.

My work focused on:

Investigating the use of a pre-trained deep convolutional neural network for image registration in 3D ultrasound prostate images. This study demonstrated the CNN's accuracy and reliability over manual and intensity-based registration methods, as detailed in the paper published in the journal Technology in Cancer Research & Treatment.
Evaluating the Clarity Autoscan ultrasound monitoring system for prostate motion tracking during radiotherapy. The research showed the system's effectiveness in tracking intrafractional motion, as outlined in the AAPM Medical Physics journal.
Assessing changes in Quantitative Ultrasound (QUS) parameters during prostate radiotherapy, suggesting their potential use in treatment planning and predicting outcomes. This research, also is detailed in the AAPM Medical Physics journal.

These advancements contribute to more accurate, effective, and tailored prostate cancer treatments, directly benefiting patient outcomes.