yash chaudhary

I am a machine learning engineer focused on building end-to-end ML systems, real-time and batch training pipelines, and scalable recommendation platforms. I design, deploy, and operate services that power personalization at scale.

At Amazon Music I built an AI agent to convert natural-language experiment specs into validated configs, automated notebook-to-production upgrades, and performed root-cause analysis for pipeline issues. I also built and ran an end-to-end training service using AWS Glue, SageMaker, and Step Functions that trains and deploys hundreds of models daily.

My work has supported high-throughput recommendation inference (20k calls/second), achieved over $200K annual AWS cost savings through optimization, and delivered a model repository on S3/DynamoDB with immutable versions, lineage, and rollback gates. I have migrated core recommendation networks from MXNet to PyTorch and led on-call efforts to ensure availability.

I bring experience designing feature extraction pipelines processing millions of time series, developing statistical ranking models used across large company datasets, and contributing to core production systems in prior roles. I seek roles where I can continue building robust, production-grade ML platforms and mentor teams on best practices.