Nothing is claimed without evidence
It is the product’s rule and ours too: no score leaves without the fragment behind it, and no evaluation closes without being fully reconstructable.
About
We built the tool we wish we had been given.
It starts from a simple idea: how someone solves a problem says more than the answer they land on. Adaliza makes that observable, without intruding on anyone.
Why it exists
We all went through technical hiring processes that measured something else: exercises nobody repeats at work, interviews where talking well outscored thinking well. Language models finished hollowing out what was left.
On the other side the problem is the mirror image: technical teams spending senior hours in interviews only to find out, right then, how the person in front of them reasons.
Adaliza moves that observation to the start of the process and records it: the candidate works an incomplete case by talking to an agent playing a role, and what stays is how they got there.
It began in an academic setting, out of research into how technical talent is assessed and why the usual methods measure the result and never the reasoning.
How we work
The decision belongs to people
We are not building an automatic filter. Everything we design leaves the decision with the hiring team, even where automating it would sell easier.
The candidate is a user too
On the other side is someone with a clock running, no training, no second attempt: that screen has the product’s smallest margin for error.
The team
Five people across product, design and engineering. Every decision in the product went through this table.
Carolina Sol Castellucci
Julia D’Alessio Poisson
Matías Pereda
Matías Ezequiel Hermida
Rodrigo Nicolás Putrino
Adaliza
The name
Adaliza joins two names from the history of computing. It is not wordplay: they are the two exact halves of what the product does.
- AdaFor Ada Lovelace
- In 1843 she wrote the first algorithm meant to be executed by a machine, and was the first to grasp that a computer could work with far more than numbers. This is the half that evaluates: criteria defined in advance, and arithmetic that applies them the same way every time.
- lizaFor ELIZA
- The first conversational program, built at MIT between 1964 and 1966. It held a dialogue by playing a role, which is exactly what the agent does in every simulation. This is the half that talks.
A rigorous algorithm and a conversation that plays a role. Adaliza is both at once, which is why it carries both in its name.
We evaluate talent. We power what comes next.
Want to see it working?
The fastest way to understand Adaliza is to see it on a role you have open today.