Sri hari Sirisipalli

I am an AI Systems Engineer focused on building scalable machine learning systems, retrieval architectures, and scientific AI applications. My work sits at the intersection of ML infrastructure, simulation-driven modeling, and production AI systems.

Currently, I develop embedding-driven retrieval and LLM evaluation pipelines used for large-scale patent similarity analysis. I built high-throughput embedding systems capable of supporting 1,000+ semantic evaluations per day and re-architected compute workloads using AWS Lambda to improve batch performance by up to 7× while reducing infrastructure costs.

Alongside this, I work on digital twin modeling and simulation-driven machine learning. I developed surrogate models trained on 88,000+ offshore simulation scenarios for motion and fatigue prediction, running thousands of ML experiments and over a million model fits to identify optimal architectures. I also built automation pipelines for engineering simulations using ANSYS and MOSES to generate structured datasets for ML training.

My broader experience includes vector search systems (FAISS), large-scale embedding pipelines, anomaly detection on sensor time-series data, and retrieval-augmented generation systems for technical document corpora.

Previously, I worked on distributed data pipelines using PySpark, optimized ML inference via ONNX conversion and quantization, and implemented monitoring systems for distributed services.

I am particularly interested in building AI infrastructure and systems that accelerate scientific discovery, digital twins, and simulation-driven engineering. My goal is to work on problems where machine learning, computation, and physical systems intersect to solve real-world engineering challenges.