Most AI projects don't fail at the model stage , they fail long before that and that's the stage I care about.
I'm a software engineer and for the past year I've been building and shipping real, production-ready web systems which covers everything from the interface a user sees down to how the database is structured underneath it.
Now, I'm turning my attention to a question most people skip too quickly when building with AI: is this data actually usable and does the approach being used even make sense for it.
Where my attention goes:
- Building complete web applications, front to back
- Applying Python and machine learning to real problems
- Getting hands-on with large language models
- Digging into data quality before anything gets trained on it
Somewhere along the way I picked up a simple rule: doing something quickly and doing it correctly aren't the same goal. I hold both my code and my AI work to that same rule.
