Mar 2026 – Present · 1 mo

• Lead a science team focused on large model training methods to advance performance across tasks and domains. Drove the launch of Model Customization and Reinforcement Learning on Amazon Bedrock, bringing fine-tuning and RL-based adaptation capabilities to production for AWS customers. Research contributions span large model training efficiency, small language models (SLMs), model training methods, agentic systems, inference optimization, evaluation methodology, and alignment techniques — all directly connected to AI products and capabilities shipped on AWS. Contributed to efforts around the Amazon Nova family of models and the Humanity's Last Exam (HLE) benchmark, with work touching foundational model training, evaluation, and applied agentic systems. Author of multiple books and patents, published across top venues including Nature, NeurIPS, ICLR, KDD, and AAAI.

Jun 2025 – Feb 2026 · 8 mos

• Led scientific and technical contributions to the launch of Amazon Bedrock AgentCore Runtime, a managed execution environment for securely deploying, scaling, and operating AI agents in production. Architected and built an internal benchmarking system to characterize service performance across latency and throughput dimensions, enabling rigorous comparisons with competing infrastructure and directly guiding engineering optimization. Drove customer onboarding at launch across integration patterns including MCP server tooling, CloudFormation-based IaC deployment, and deep research agent architectures, scaling to thousands of customers. Contributed to the AgentCore Starter Toolkit SDK to lower the barrier for developers bootstrapping production-grade agent deployments. Applied reinforcement learning techniques for customizing multi-turn agentic systems, advancing the scientific foundation underlying the platform's agent execution model. Enabled production workloads spanning compliance screening at transaction scale, agentic RAG, and autonomous deep research pipelines.

Oct 2022 – May 2025 · 2 yrs 7 mos

• I interface with science leaders and research teams on the customer side, guide customers to their best course of action when implementing state-of-the-art algorithms and techniques, and collaborate with Amazon and customer teams in high-impact research directions leading to direct business benefit.

Oct 2020 – Oct 2022 · 2 yrs

• I help AWS customers build and fine tune large-scale Machine learning and Deep Learning models, and rearchitect solutions to help improve security, scalability, and efficiency on the cloud. I have helped customers set up secure Machine Learning platforms for large teams, and specialize in setting up massively parallel distributed training, hyperparameter optimization and reinforcement learning solutions. Apart from Machine Learning, I also help solve complicated distributed computing and optimization related use cases.

Apr 2020 – Oct 2020 · 6 mos

Jun 2018 – Apr 2020 · 1 yr 10 mos