Machine Learning Engineering
Model development, evaluation and deployment. Classical ML, deep learning and the MLOps glue that keeps them honest.
- Training & eval pipelines
- Model serving & monitoring
- Experiment tracking
Software and machine learning,
built for production.
Cross Step Labs is an independent engineering studio. I help teams ship production ML systems, data platforms and the product software around them — from first prototype to something you can trust at scale.
Illustrative data · Seasonal retail
Services
Where science meets engineering.
Two decades of turning models into production systems, from research prototypes to platforms serving hundreds of millions of predictions.
LLM & AI Applications
Retrieval, agents and structured generation that work on real data — with evals so you know when they don't.
- RAG & search
- Agentic workflows
- LLM evaluation suites
Software & Data Engineering
APIs, backends and data pipelines built with reliable tools. Clean code your team can own long after the engagement ends.
- Python, TypeScript, Java
- Streaming & batch pipelines
- Cloud infrastructure
Selected work
Recent projects.
Highlights from the last few years.
Demand forecasting at Amazon scale
Launched a deep learning forecasting model covering more than 400M products sold on Amazon, improving forecast accuracy by over 1,000 basis points.
Amazon Now
Launched demand forecasts for Amazon Now, enabling sub-30-minute delivery of grocery and household items.
About
Hi, I'm Shigeru.
I'm a software and machine learning engineer with over 20 years of experience. Most recently I spent 12 years at Amazon, finishing as a Principal Engineer and chief technologist for the Forecasting organization consisting of 180 engineers, scientists and product managers.
Earlier in my career I worked on the more scientific side. At MathWorks I designed probabilistic models and algorithms such as electricity price simulations and differential equation solvers. Before that I spent a year as a researcher at Carnegie Mellon through the MSE Fellowship.
I started out in finance, leading development of Deutsche Bank's fixed-income trading platform and risk calculation engine for the Asia-Pacific region.
“Shigeru has been at the center of nearly every major shift in our stack over the past decade. He led the engineering design and launch of our earliest deep learning models, set the engineering vision, and has been the engineer we turned to for almost every new initiative that came our way.”
Director of Forecasting, Amazon