About
Niitaka didn't start
as a company idea.
It started the way most internal tools do —
trying to make sense of systems that were changing faster than we could reason about.
Working on data and production systems, a pattern kept showing up. You make a change — a new model, a different prompt, a tweak to logic — and suddenly you're staring at a mix of logs, metrics, and outputs trying to answer a simple question:
Did this actually make things better?
Sometimes the answer was obvious. Most of the time, it wasn't.
Where this comes from
I work as an analytics engineer in healthcare, and previously as a data scientist at Meta. I've spent a lot of time in systems where small changes are expected to be measured carefully, and decisions need to hold up beyond intuition.
When I started working more with AI agents, something felt off. I know how to code — building agents wasn't the hard part. There are already plenty of tools that help with that: frameworks, debugging utilities, trace inspectors.
But that wasn't really what I needed. The harder part was knowing what to trust.
You change a prompt. You adjust a workflow. You try a different model. It's easy to make something look better. It's much harder to prove that it actually is.
I remember looking at two versions of an agent — both seemed fine, metrics were slightly different — and still not being able to confidently say which one we should ship.
From an engineering perspective, that felt unacceptable. Because at some point, you still need to answer:
Is this version actually better?
Is it stable enough to ship?
Can I explain this decision clearly to someone else?
That gap became more obvious over time. The tools existed, but they were solving a different problem.
They helped you build.
They helped you debug.
But they didn't help you decide.
Niitaka comes from that experience. Not from trying to build another observability tool, but from wanting a clearer, more structured way to answer a practical question:
When something changes, how do you know if it's actually an improvement?
Today, Niitaka is an attempt to make that process explicit — to take what usually lives in scattered logs, intuition, and one-off analysis, and turn it into something you can reason about, compare, and act on.
We're building this in public, and the shape of the problem is still sharpening. But the goal has stayed the same since the beginning.
Measure what changed.
Decide what ships.
On the name
Niitaka (新高) is Japanese for “new heights.” It felt right for a tool built around the idea that every change should be measurable — that better isn't just a feeling, it's something you can demonstrate.
Try it yourself
Niitaka is free to start. Connect your agent in minutes and see what's actually changing.