Turning Prior AI Projects into Practical Service Research
Consulting and data-marketplace research explores how concrete project workflows might become narrowly defined services.
That means I’ve been hands-on since the very beginning of the modern wave—starting with early autocomplete and chat tools, and moving right along as frontier models, agent harnesses, and APIs matured. I’ve been dabbling in data science and machine learning foundations since 2015, which makes today's accessible LLM era feel like an incredible leap forward from earlier, rigid modeling workflows.
Now, LLMs and agents are woven into everything I do. I use whatever model or tool makes sense for the job across the command line, desktop apps, and web interfaces. Because my background is in product, operations, and data, I don't just chat with models—I build with them. I wire AI features into apps and set up agent workflows, which means figuring out how to give them the right context, tools, and checks so they can actually help plan, code, test, and troubleshoot without going off the rails.
Away from writing software, I use these systems as everyday leverage: researching strange problems, analyzing data, working through decisions, and hooking them into my files, emails, calendars, and desktop workflows to handle the heavy lifting.
Consulting and data-marketplace research explores how concrete project workflows might become narrowly defined services.
A paused local dashboard project explores event accounting, duplicate records, and the limits of token-based usage summaries.
Remembering Sir Patrick Stewart through his lockdown Shakespeare readings and revisiting the #ASonnetADay dashboard I built to help fans find them.
Citrini argues for realistic medical timelines. AI can still accelerate discovery by improving planning, analysis, and the organization of research.
Black October's security warnings call for practical action: patch software, isolate VPS servers, protect accounts, and make time for regular reviews.
A product-video workflow combines approved narration, cached speech, captions, and screenshot-based rendering.
David Robinson is right to flag OpenAI's safety failures, but the incidents he describes call for concrete engineering controls.
A photo-navigation experiment separates a working viewer from optional model inference and reconstruction ideas.
A product specification explores turning fragmented personal traces into semantic activities and recurring day patterns.
My NASA Virtual Guest passport now spans five mission stamps, while my name is flying to Jupiter aboard the Europa Clipper spacecraft.
A reassessment of Gestalt communication in Diaspora and how generative AI could help people align complex mental models with less explicit explanation.
A scheduled reminder-call service explores deterministic speech, outcome evidence, account separation, and billing.
Counts show the breadth of the writing without turning every page into a category wall.