Deep technical leadership
I lead from first principles, stay close to the code, and make architecture decisions with enough depth to earn trust from both engineers and stakeholders.
Director of Engineering, Applied AI
Systems that think. Interfaces people can use.
I’m Dakota Kim, a deep technical leader working across applied AI, product engineering, and human-computer interfaces. I lead teams, design systems, and ship software myself, moving comfortably from architecture and native apps to language models, agent workflows, and research-driven tooling.
Approach
Technical depth with a human outcome
The throughline in my work is consistent: go deep enough to understand the system, then shape the product so that technical capability turns into something genuinely useful for people.
Applied AI with product discipline
I care about reasoning systems, agents, and AI tooling, but always through the lens of reliability, usability, and what actually helps someone do better work.
Professional, with a pulse
I take the work seriously without sanding off the personality. Curiosity, care, and a bit of play are part of how I build products people remember.
Work
Selected work across products, research, and tools
This is the more complete picture: leadership in applied AI, a growing body of independent research and open-source work, and a product portfolio that spans web, desktop, and Apple platforms.
MadWatch
An independent software lab where I build thoughtfully crafted tools, apps, and services across platforms, with an emphasis on usefulness, care, and strong product taste.
reasoning.software
An applied research arm focused on reasoning systems, world modeling, human-computer interfaces, and the practical edges of AI systems.
Hugging Face research releases
Published models and experiments exploring fine-tuning, reasoning, and agentic systems, including active recent work under GhostScientist.
semanticwiki
An AI-powered architectural documentation agent and coding assistant built to generate traceable technical documentation with real code awareness.
Copyist
A lightweight clipboard utility for macOS that reflects my bias toward practical software with a clear job and a clean interface.
Daily Quest
A habit product framed as gentle, long-term quests, showing the softer product language and motivational design choices I’m drawn to.
The Tiny Met
A watchOS app that brings the Metropolitan Museum of Art to Apple Watch, pairing cultural exploration with compact interface design.
HackerWatch
A Hacker News client for Apple Watch, built as a fast, focused experience for staying connected to the tech conversation from your wrist.
Writing
Notes from the work
I write about AI tooling, engineering control surfaces, agent workflows, and the kind of curiosity that tends to leak across disciplines. The writing is technical, but it still sounds like me.
Agent Skills Work for Humans Too
How patterns designed for agents can sharpen human workflows and learning loops.
The Levers of Control in Claude Code
A practical look at what makes AI coding systems predictable, steerable, and useful.