Prototyping input interactions for next generation AR Glasses on the Reality Labs Input Interactions (IX) team.

Working on creating the first version of a hand tracking input system in Facebook Horizon, an online multiplayer virtual reality game developed for Oculus Quest. Implemented a system for detecting hand tracking gestures as well as a hands-only locomotion system utilizing teleportation and snap turns.

I am currently working on an alphanumeric classifier and a method to calibrate machine learning models for the output of valid confidence levels. The Vision software subteam focuses on the Automatic Detection, Localization, and Classification (ADLC) of targets imaged from a high-altitude autonomous aircraft.

• Developed end-to-end workflow for training and evaluation of face clustering classification models, reducing the frequency of costly reclustering by 60-70%.

• Fully automated classifier evaluation on custom data set partitions, integrated with CLI tools deployed to production.

• Extended the APCV Experimental Platform to visualize classifier evaluation results.

• Piloted conversion of product classification to an NLP based, automated system.

• Utilized the ULMFiT transfer learning method, with a top accuracy of 97% in testing.

• Created new product classification taxonomy and organized the retagging of ~50,000 products.