Sudhir Pol

Machine Learning Engineer with 3+ years of experience building production machine learning systems at Adobe, S&P Global, and American Express. I work across agentic AI, LLM evaluation and serving, NLP, and MLOps, from prototyping and fine-tuning to deployment and observability on AWS and Kubernetes.

Earlier roles at S&P Global and American Express included document intelligence, production NLP services, and calibrated risk models used in live business workflows.

What I am looking for

I earned an M.S. in Data Science from Indiana University in May 2026 and am seeking full-time roles in the United States in machine learning engineering, agentic AI, LLM systems and model serving, or applied data science. I am especially interested in teams that turn research ideas into reliable products and measure their impact on quality, latency, cost, and user outcomes.

Current technical interests

My current technical focus is dependable agent architecture: explicit workflow state, MCP-based tool interfaces, human-in-the-loop controls, and end-to-end tracing. Alongside this, I work on CUDA attention kernels, quantization-aware routing, speculative decoding, model evaluation, and high-throughput serving.

Selected writing

I write implementation-focused articles on quantization, speculative decoding, and LLM inference. They are collected on the writing page.