Models are part of the story.
The system is what makes them useful.
My areas of work span local inference, agent orchestration, retrieval, evaluation, optimization, and multimodal production applications.
I work on running language models locally and understanding the practical choices around inference. The questions that interest me are how to balance quality, memory, context, and speed for the work at hand.
I am interested in the systems around a model: planning, tool use, state, and the checks that make multi-step workflows understandable. The goal is useful automation without losing sight of what happened.
Retrieval systems connect a question to relevant information. I focus on making that connection useful and inspectable, with attention to the information being retrieved and the answer it supports.
Evaluation and optimization are connected parts of my practice. I want to understand how a system behaves on the actual task before deciding which model, runtime, or workflow is the right fit.
I work toward production applications that can bring text, images, documents, and other inputs into a coherent workflow. The challenge is turning model capability into something understandable and usable.
My direction is forward-deployed engineering: understanding a domain, building with the people doing the work, and carrying the solution from an early experiment toward an application.
Finance and healthcare are my primary focus.
Robotics and physical systems are an emerging interest.
Exploring how language models, retrieval, and agents can make financial information more accessible, structured, and useful.
An area of focus, not a claim of deployed financial products.
Interested in knowledge retrieval, multimodal information, and workflow tools that support healthcare professionals without obscuring their judgment.
An exploration area. No clinical tool or medical claims are presented here.
Following the connections between AI, robotics, and the mechanical, electrical, and structural systems that make real-world machines possible.
A learning direction, not an established robotics portfolio.
“Start with the problem.Explore the planned experiments
Make the system understandable.
Let the evidence guide the next step.”
From a first question to a system in the real world.
Let’s find the right place to start.