- Leading the Semantics team within the Perception & Prediction org- The team is contributing to the full AI cycle from data acquisition, model training, evaluation, to on-car deployment- Solving hard problems in autonomous driving and machine learning such as (i) Traffic light detection and state estimation; (ii) Special actors (Emergency vehicles, Traffic controllers etc.) detection and activity recognition; or (iii) Visual semantic features for Prediction- Supervising and up-leveling team of 7 ML engineers on topics including the SOTA/best-practices in Deep Learning/ML

● Three promotions in four years, from SDE2 to Principal level.● Tech lead for various Machine Learning / Computer Vision product and research efforts with teams of up to 10 contributors. This involved mainly internal and external engagements from customer communication to problem specification, execution, testing, and deployment.● Worked on wide range of problems including classification, regression, clustering, similarity, ranking, etc. using either tradition Machine Learning (e.g. Random Forests, SVMs) or Deep Neural Networks.● Supervised and mentored over 25 people during weekly 1:1s in addition to working as independent contributor.● Started and grew the PyTorch Computer Vision Best Practices repository to over 8,000 stars on GitHub: https://github.com/microsoft/computervision-recipes ● Presented work at conferences such as the O'Reilly AI conference, the Open Data Science Conference, or the Microsoft's AI Immersion workshops.● Participated in internal culture efforts including diversity & inclusion or team morale. ● Wrote a C# cell phone app in my free time to teach how to read notes called "Piano Teacher” and published the app to the windows phone store.

● Contributed to Microsoft's Natural Language Processing (NLP) experiences for Bing and Cortana.● Experienced full product life cycle, from architecture to implementation, evaluation, testing, deployment, monitoring and maintenance. Mainly used C# and Python.

Researched and wrote production code in Machine Learning areas such as face recognition, facial feature detection, intrinsic imaging, or chromaticity estimation. Mainly used C/C++ and Matlab. Two patents granted, one of which first author.

Taught and graded multiple undergraduate classes on a weekly basis.

Developed a real-time Machine Learning solution for eye blink sleepiness estimation in C/C++ and Matlab. This work was carried out in a team of Japanese researchers, where I was responsible for design decisions, software implementation and evaluation.