Priya Ramani
AI Infra, Pytorch @ Meta | Ex-OCI, Ex-Zoho
- Role
- Senior Software Engineer & Product Owner at Meta
- Location
- Seattle, WA, US
- LinkedIn followers
- 500 followers
Experience
Senior Software Engineer & Product Owner
May 2021 — Present · Seattle, WA, US
Leading and building the industry leading Artificial Intelligence framework, Pytorch and its components- Expertise in multi-folding model processing efficiency including preprocessing, training and inference of models- Cross platform expertise with building TorchX for Pytorch on Google Cloud (GCP Batch)> Pytorch Recsys Preproc- Current building Pytorch preprocessing frameworks, to power efficient preprocessing of data for Recommendation systems- Developing and optimizing GPU and tensor-based operators, enabling the migration of real-time model data preprocessing from CPUs to GPUs across Meta’s Ads stack- Enabled preproc for long sequence event based data with sophisticated mechanisms for data representation, and performance optimization- Validating correctness is one of the core themes critical for AI assessment in the current world. Developed strategy and comprehensive multi-layer testing tools for correctness and performance benchmarking.> TorchX- TorchX - GCP cloud integration: https://cloud.google.com/blog/products/compute/rapidly-deploy-pytorch-applications-on-batch-using-torchx - Lead TorchX integration with Google Cloud Batch to enable scheduling, training, and developing AI models and Pytorch applications on Google Cloud. Provides faster setup, development, iteration and productionization time.> Pytorch Edge (OnDevice)- Model compilation for OnDevice efficiency: Built compiler driver, shape propagation mechanisms and workflows for compiling, validating, iterating, integrating and delivering compiled models to mobile. This enabled Integrity models run faster with reduced app size- Performance efficiency of AI enabled smart devices: Optimized the time to action on smart watches. For eg, the time between when you say “take a picture” and device taking a picture. built light weight static dispatch on multiple backends in Pytorch to which reduced pytorch initialization time by 45%.
Education
Anna University Chennai
Bachelor of Engineering (B.E.), Computer Science
2010 — 2014
Skills
- Public Speaking
- Javascript
- Type Systems
- C
- Go
- Data Modeling
- Language Development
- Compiler Construction
- Java
Find verified contacts for anyone on LinkedIn
Unifers gives sales teams verified emails and direct dials, enriched profiles, and outreach that lands in the inbox.
Free plan included · No credit card required
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.