Building real-time AI inference pipeline for Meta Superintelligence Labs. Working on performance optimizations, scheduling, batching and orchestrating, for LLM, DiT and other model architectiures

Focusing on bringing PyTorch (runtime, interpreter) into embedded systems, mobile phones, desktops and laptops. Improving its modularity/scalability so that it satisfies different hardware constraints while supporting PyTorch model inference on devices.

Working on Facebook Stories, building new infrastructure for publishing & presenting stories on Android platform.

Development on Oracle Cloud Machine. Rich experience in Juniper/Cisco switch configurations, implemented a tool to automate the configuration process. Developed health check tools that currently are being used in OCM, both projects are written in Python.