Senthilkumar Gopal

Musings of a machine learning researcher, engineer and leader

About me

I am a machine learning researcher, technologist and an avid engineer focused on identifying and building creative solutions and teams to solve for complex problems. I had the good fortune of working on various technologies from highly scalable and available web based products, end to end services and architecting frameworks, machine learning algorithms and currently learning hardware-software co-design for ML acceleration.

I am currently leading the ML Framework Extension (for PyTorch, TensorFlow and JAX) for Inference acceleration using AWS Trainium and Inferentia Accelerator chips managing team of machine learning engineers driving critical projects for internal (Bedrock, Alexa) and external customers for onboarding their deep learning models and support several large Enterprise customers on Neuron based accelerator instances. I am responsible for adoption of leading LLM models such as LLamav2, Mistral, GPTJ and other Gen AI models such as GPT-NeoX and BLOOM, saving training and inference costs over comparable GPU-based EC2 instances. I am driving NeuronX Distributed library for customers to parallelize large models across training and inference tasks using tensor/sequence and pipeline parallelism.

I received my Masters in Computer Science from Georgia Tech with specialization in Machine Learning.

My recent work and research in recommendation systems involved comprehending user provided content, specifically around unstructured item and inventory information aiding in improving information retrieval for eBay search. My work focused on building and tuning ML models to understand and qualify item and inventory data to assist in improving search quality, create feedback loop for sellers to understand the importance of their content and drive for clarity and conversion, and build new experiences and workflows in eBay. My primary focus was on applying modern machine learning technologies and research to improve the comprehension of Item/Inventory and build semi-supervised and supervised ML models for various content related work around aspects understanding and quality, image quality controls and coherence of various measures around item inventory.

Earlier in my career, I developed multiple Identity Infrastructure products such as Identity Linking Platform, OAuth management, Secure Token Authentication systems and Federated Identity which form the critical platform components of flows such as checkout, search etc., in eBay with focus on security and availability. I also built end to end solutions for integrating social identity providers in eBay ecosystem, creating a single use code authentication platform, redesigning authentication tokens data model and overall Identity architecture. During this period I also worked on PCI DSS compliance, Encryption standards, OAuth standards, OpenID Connect protocol and intelligent ATO prevention and bot attack mitigation measures.

I worked on multiple designs and development of Identity Architecture for eBay building Authentication platform products, next generation of API authentication protocols, Identity enablement systems, their data architecture and other elements of the authentication ecosystem.

During my tenure as the engineering leader in Toronto, I managed multiple teams for integrating Terapeak Research Analytics platform into eBay marketplaces. The multi fold architecture accomplishes an end-to-end data processing pipeline and application architecture for loading data and to garner better insights to sellers on a near real-time basis.

Subsequently I was managing the Seller Marketing platform at eBay and leading the engineering teams responsible for one of the most critical components of eBay marketplace which has touchpoints all over the seller and buyer experiences including search, view item, checkout, orders etc., including the most recent product release of Seller Coded Coupons. This platform had a very wide range of capabilities for sellers and buyers for promotion lifecycle management, checkout and cart redemption and the whole end to end pipeline for promotional integration in ecommerce.