Felipe Sanz — Applied AI Engineer & Data Scientist

Hi, hola, salut,

My name is Felipe Sanz and I'm a Data Scientist and Applied AI Engineer based in Austin, TX. I grew up in Colombia, moved to Switzerland for college and an early career in finance, then lived in Madrid, Spain for 4 years before landing in Austin.

I have worked with data my entire career, moving across the stack from data capture to analysis to machine learning model development. Most recently, I have worked on designing useful AI systems that deliver tangible and measurable results. This includes AI memory architecture, observability, and evaluation.

I deeply believe there is a mismatch between how AI is being adopted now and how it should be adopted if we want it to act in favor of humanity and not to its detriment. This is what I write about: an opinionated view of how to create AI for good.

The best way to follow my activity is and .

I hope you find something here worth your time.

I write every article in English, Spanish and French. Use the globe icon above to toggle between languages.

Latest articles

A cardboard-craft still life on white paper: a small paper graduate in a mortarboard stands where a cardboard path forks. To the left, the path turns dark grey and breaks into scattered fragments that disappear under a torn sheet of cream paper. To the right, three continuous teal lanes widen and sweep away into the distance.

The AI hiring split:
my playbook for new grads and companies

The entry-level squeeze happens through jobs that never get posted, which makes the damage almost invisible to layoff trackers.
Read More
A cardboard-craft still life: on the left, a spool of dark cable labeled '.com' with loose cables and scattered cardboard microchip tiles; on the right, a stack of cardboard server-rack units labeled 'AI' — the dot-com era set beside the AI buildout.

AI vs dot-com:
where they rhyme and where they don't

The AI buildout is not just a bet on demand. It is a bet that compute remains scarce long enough to repay the cost of building it.
Read More
A cardboard-craft still life: a conveyor of paper payroll cards, each stamped with a worker icon and a dollar sign, feeds into a central press that outputs a row of black server-rack blocks — payroll being converted into compute.

Payroll for Compute

For the biggest technology companies, the most important AI-layoff mechanism is often capital allocation, not direct labor substitution.
Read More