nikam.dev ← All cases Telegram →
nikam.dev / Cases / Astrolab - career-guidance platform
Live · career guidance · teens + adults

Career guidance grounded in a method, not a guess.

A service that helps a teenager choose a profession and an adult change one, grounded in a real method instead of a 'which celebrity are you' quiz. You take a test (interests, school subjects, values), get a visual profile and professions in three baskets - from the obvious to the unexpected - and for each one you immediately see how to actually get there: field of study, exams, deadlines. The test needs no signup. AI explains the match but never invents professions. Teens work end-to-end today; adult career-change is the next contour. Astrolab is a non-commercial social project: completely free, no ads, no data trade. It's already live in three languages - Russian, English and Spanish - with more on the way.

What
Career-guidance platform
For
Teens + adults
Method
RIASEC + Klimov + values
Status
Live · free · RU/EN/ES
Open astrolab.nikam.dev →
— Problem

People pick a profession blind - and the cost of a wrong pick is measured in years.

A teenager decides who to be based on parents, prestige, and a random subject that 'goes okay'. An adult sits in the wrong profession for years because switching feels like starting from zero. And the tools around them either entertain, or scare with a price tag, or hand back a single 'right answer'.

Pain 01
Chosen by parents and prestige, not by the person
'Go be a lawyer or an economist, it's safe.' The decision is made for the teen and without any data about them: what they enjoy, what comes easily, what matters to them in work. A few years later - disappointment, lost time, and money spent on the wrong education.
Pain 02
Tests are either toys or a black box
Some career quizzes are 'which superhero are you', others are paid and opaque: it's unclear where the result came from or why this profession. Zero trust, and the decision is for life.
Pain 03
The profession is clear - the path in is not
Say the direction is chosen. But which field of study, which exams to sit, what deadlines to file by - that's a separate quest across dozens of sites. Between 'I want' and 'I got in' stands a wall of unclear steps.
Pain 04
Adults can't see that their experience transfers
Someone wants to switch fields but fears their whole career will reset to zero. In reality part of their skills carry over into the new profession - only nobody shows it clearly, so people stay where they feel stuck.

A fifteen-minute test - and a clear map: who to be and how to get there.

Not guesswork and not 'AI magic'. A transparent method computes the profile, professions are matched from a vetted catalog, and each one comes with an honest answer on how to get in. It all runs live on astrolab.nikam.dev - test, profile, professions, and the admission path.

A test on method, not on mood
Interests (cards in RIASEC logic), school subjects paired as 'like x good at', work values - adjusted for age and stage. The question bank and scoring rules are versioned: the result is reproducible, not 'however it landed today'.
Your profile shown, not stamped
A RIASEC radar and a values profile instead of the verdict 'you're a humanities person'. You see the picture of your inclinations, not a single label pinned on you. The signature radar is a recognizable part of the product.
Professions in three baskets
'Core' is a direct hit, 'Nearby' is adjacent, 'Dark horses' are unexpected but on-profile. The output doesn't collapse into one answer: you see the obvious and the things you'd never think of yourself.
AI explains the match - but never invents
For each profession the AI writes a warm, honest 'why you' and what to watch out for. But it only re-ranks and explains - it cannot add a profession outside the vetted catalog. No hallucinated 'perfect dream job'.
Every profession is its own page
What the job involves, which inclinations it needs, a salary reference honestly badged 'estimate'. SEO pages with structured data: professions are found from search, not locked behind the test.
'How to get in' for Russia, honestly
Fields of study (official codes), typical exam combinations and admission deadlines - curated by hand and marked 'verify on the university site'. For other countries, an honest 'no vetted data' plus links to official sources, with no invented paths.

Honesty is wired into the architecture, not into a disclaimer.

A product about a life-shaping choice has no right to hallucinate and no right to hoard extra data about children. So the constraints live in the code: what must not be faked, technically cannot be.

/ 01
A method you can't quietly tweak
RIASEC + Klimov mapping + subjects + values. The scoring rules are immutable: the scoring config is pinned in the database and protected from edits at the DB level - without it the test simply won't start. The result is reproducible and versioned.
/ 02
A formula computes, AI only explains
A deterministic engine computes the profession match (interests, subjects, values). AI re-ranks the top and writes the explanations, but adds no professions - anti-hallucination is built in. Every model call is logged.
/ 03
Anonymous-first - no child data piles up
The test is taken without signup; the result lives on an expiring token link, and an account is only needed if the person wants to save it. Minors' personal data isn't collected by default - and conversion is higher for it.
/ 04
Every fact has a source
Data carries provenance, a date, and a confidence level; salaries are honestly badged 'estimate'. The profession catalog passes a panel of ten AI models rather than a rubber stamp: disputed entries are held until reviewed.
/ 05
The result never depends on AI
The match is returned instantly by the deterministic engine, and AI explanations enrich it afterward. If the model is unavailable, the person still gets a result, just without the text notes. AI assists but never blocks.
/ 06
One backend, an SEO catalog, deploy in one push
FastAPI + Postgres + SvelteKit SSR, an i18n architecture built to grow, LLM via my own subscription (zero variable cost at MVP). Push to the repo, live in production within a minute.
— What already works

From idea to a live product - the full cycle in production.

4 cycle steps
Test -> profile -> professions -> how to get in run end-to-end on astrolab.nikam.dev: a teen walks the whole path, not just a bare list.
0 invented professions
The matching is deterministic, AI only ranks and explains. A product about a life-shaping choice has no right to hallucinate - and architecturally cannot.
0 signups for the test
Anonymous-first: a test with no account, a result on a link. Minors' data isn't collected by default - privacy as a decision, not a checkbox.

How it looks.

The English build - landing, the US professions catalog, and a profession page with the RIASEC radar and an honest admission path.

Astrolab English landing
Landing - how it works, what you get, and the honest approach
Professions catalog with filters
Professions - a searchable US catalog with pay and education filters
Profession page with RIASEC radar
Profession - RIASEC radar, a real BLS salary, and a US admission path

Have an idea for a product with a method at its core?

Astrolab is my own non-commercial initiative - a free social project taken from method to a live product, now running in three languages. If you have EdTech, an assessment tool, or a consumer service where transparency, privacy, and 'AI that doesn't lie' matter - in 30 minutes we'll map how to build it honestly and what to show first.

Message on Telegram → nick@nikam.dev