What if a student could ask “what job am I actually ready for?” and get a useful answer — not a brochure, not a generic aptitude quiz result, but a real answer backed by 500 schools of curriculum data and live signals from the job market?
That was the question SENAI needed to answer. We built the engine to answer it.
The Black Box Nobody Talked About
SENAI trains millions of professionals across Brazil’s industrial sector every year. The scale is genuinely impressive: 500+ schools, a national curriculum catalog, a workforce development mandate that touches nearly every industry in the country.
But there was a gap nobody had figured out how to close. On one side: students who wanted to know which training would actually get them hired. On the other: employers who needed talent with specific skills. In between: SENAI’s curriculum — thousands of courses, itineraries, and learning paths — with no systematic way to match any of it to what the market was asking for right now.
Career guidance at SENAI, like at most institutions, was largely a human judgment call. A counselor would meet with a student, review their background, and recommend something from the catalog. Well-intentioned. Not scalable. And based on whatever the counselor happened to know about the job market that week.
We set out to make the invisible logic explicit.
The Dating App for Your Career
Think of it like a dating app — except instead of matching you to a person, it matches your profile to a job. And instead of photos and bios, the matching engine reads three different data sources simultaneously:
- Your professional profile — declared skills, education history, work experience, what you already know
- Live job market signals — what employers are posting, what skills are in demand in your sector and region right now
- SENAI’s National Itinerary — the full catalog of available courses, certifications, and learning paths
The engine compares all three and produces something specific: an employability probability score. Not “you should learn Python.” More like: “Based on your current profile and market demand in the TI-Software sector, you have a 75% probability of qualifying for this role. If you complete this course, that rises to 82%.”
That’s a different kind of answer. It’s actionable. It’s personal. And it makes the implicit logic of career matching explicit enough to trust.
“For the first time, a student could see not just what to study, but why — and what job was waiting on the other side.”
What the System Actually Did
We built the AI MVP in partnership with Google and Atos, initially scoped to the TI-Software sector and one job portal. The pilot gave us a controlled environment to validate the core logic before expanding.
The engine’s outputs weren’t just student-facing. We also built an institutional dashboard — an employability observatory — that surfaced aggregate data across regions and sectors. Which skills were most in demand in São Paulo this quarter? Where were the biggest gaps between what SENAI was teaching and what the market needed? That view was new information for SENAI’s curriculum planners, and it changed how they thought about program design.
🚧 Need more context: Which specific AI models or APIs from Google and Atos powered the matching engine? How many students were processed through the pilot? Do we have production adoption data or employer partnership outcomes from the rollout?
Why This Was Hard
The technical challenge was real — building a matching engine that could reason across three heterogeneous data sources, output probabilistic scores, and do it at national scale is not a weekend project.
But the harder challenge was epistemic. For the system to be useful, the job market data had to be current, the skill taxonomy had to be consistent, and the curriculum catalog had to be structured enough to be machine-readable. None of those things were in perfect shape when we started. A large part of the work was making the data coherent enough for the AI to reason over.
That work rarely shows up in product demos. It showed up in the accuracy of the scores.
What It Meant
Career guidance has always been limited by the information available to the person giving it. A good counselor can only know so much about the current market. The AI engine didn’t replace that counselor — it gave them something they never had before: structured, current, personalized data to work from.
Students who used the system didn’t just get a course recommendation. They got a career path: here’s where you are, here’s the gap, here’s what fills it, here’s the job on the other side. That’s a fundamentally different product than a course catalog with a search bar.