Deepti Chheda

* AI Personalization: Building MetaSearch, Personal & Social Intelligence, and Private AI experiences powering the MetaAI assistant * Instagram Relevance - built SOTA ML/AI recommendations systems to power Instagram’s Feed, Reels, Stories, Notes ranking, for both connected & unconnected recommendations . These drove a majority of overall IG time spent and engagement metrics. - Core modeling: Built Early-stage/Late-stage ranking models that have the most complex model architectures. Innovated new dense and sparse scaling techniques, sequence modeling, longer user history length, knowledge distillation, cross-surface & cross-domain learning . - ML infra: Tackling model training, model scaling, online training, data scaling , model co-design for heterogeneous GPU hardware types, scaling the models for 2B+ users. - Delivery: App-start, latency, caching, prefetching and delivery of all content on the IG app * Content Ecosystem: As part of MRS initiative, built the ML models for FB/IG that powered Content understanding, incl. content taxonomy, content quality models, originality models, trend detection models, and content distributions models. These overall helped in promoting more original content and helping new / viral / timely content breakout more quickly in Feed/Reels recs * User Understanding: Built ML models that predict users’ age, language, location, close friends, user’s interests . These predictions are used heavily by recs, growth , integrity teams. * IG Growth: Led all Instagram Growth efforts for Instagram across acquisition, activation, friending, account recs, notifications, external sharing which are the biggest growth levers. These were vertically integrated teams that are responsible for the entire tech stack across - mobile, web, backend, ML, infra, prod infra.

Built Uber’s data, ML and recs systems from 0->1 during the hypergrowth years

I was a founding member of the startup. Built a distributed cloud file system using FUSE. Grew the company/product from 0->1, found product market fit, and later was acquired by EMC

Built a couple of 0->1 storage products, took it to market

Windows Kernel

Intern. Worked on NUMA processors

My research & passion was in the areas of storage systems, distributed file systems and performance/efficiency.