Current : AI Infra / PyTorch Team for modernizing recommendation systems at Meta using LLMs , Improving reliability and efficiency of distributed training and inference for RecSys Models using PyTorch2 , GPU / Triton.

Previous :

• AI Training Infra : Resource management and job scheduler for GenAI , RecSys jobs across 100K GPUs.

• WhatsApp Infra : Scaling services for 2.7B+ MAU at WhatsApp Messaging Infra.

PS5 Platform Architecture and I/O Stack , Application delta patching, Machine Learning for I/O compression and deduplication. Patents : 11449325, 11307841

• Lead development of Linux Kernel Block Layer caching software (called Samsung AutoCache for Enterprise Linux)

• Lead development of Linux Kernel log structured block device for improving write endurance of client SSD/eMMC.

• Lead development of Flash Aware compression based DRAM caching solution for Android Kernel.

Lead for Veritas Cluster Filesystem (VxFS) Development & Escalations. Developed OpenStack Cinder driver for VxFS.

Startup for measuring and rating "code quality" of a software.