About Me

I’m an electronics engineer who moved into software and AI engineering. I’ve worked on production AI systems involving RAG, cloud infrastructure, APIs, CI/CD, and AI-assisted workflows, primarily using Python, GCP, and modern LLM APIs.

I started FounderForesight to investigate a question I kept running into: what does it actually cost to run AI yourself?

It’s easy to compare model prices on API pricing pages. It’s harder to understand what happens when you run an open-weight model on your own hardware or cloud GPU—how much compute it uses, how performance changes with different configurations, and what happens to the economics as the workload changes.

So I’m building and publishing experiments rather than relying on assumptions. Each experiment has a defined setup, measured results, and documented methodology. I use primary sources where possible and separate what I measured from what I’m inferring.

There are also limits to this research. The current experiments focus on single-GPU systems and 8B-class models. I haven’t tested every model, hardware configuration, or production workload—and I’ll say so when something is outside the evidence.

If you’re interested in the next benchmark runs, subscribe to FounderForesight. If you’re deciding whether to build or buy AI infrastructure, you can also book a call to discuss the technical and cost questions you’re trying to answer.