Meta Reality Labs

VR Camera firmware:

Synchronizing and controlling multiple cameras in real-time, along with IR LEDs, on an MCU. Bringing up new camera sensor drivers and working with external vendors for configuration and camera initialization sequences.

OS (AOSP):

Camera power sequencing, communicating between AP and MCU to drive cameras and consume frames for downstream tracking engines.

Shipped Quest Pro, Quest 3, Quest 3s, and more to come :)

Azure Hardware: AI Accelerators

Writing C++ device firmware to implement large NLP models on new hardware, with additions to our library SDK and hardware simulator.

Applied Data and Machine Learning Scientist in Microsoft Advertising, Bing Ads

Reinforcement learning, Contextual bandits and bayesian inference; Advertiser embedding models (Triplet-loss NLP models; leveraging Transformers like RoBERTa and UniLM); Statistical analysis on advertiser churn, performance, time-series forecasting

Image-Ad relevance models — generating relevant images per ad at scale using Spark to be shown on the Bing Search page

Deep Learning NLP for Alexa (LSTM, CNN, BERT) in Keras/TF and PyTorch for profanity and hate speech detection.

Object similarity search with A-kNN on embeddings using FAISS at scale to de-dupe policy violations.

Big Data processing with Spark/Hadoop in AWS EMR and Scala for model inference.

ETL into Redshift DBs for analytics in AWS QuickSight

Java Spring Back-End -- Built API endpoints for Alexa customers to manage their live skills for deletion or suppression on a self-service basis, able to handle thousands of transactions per second.