Raghav Jalan

Jaipur, India

All work

6 of 6

Career Transition

Paths, not listings.

Status
In development, side project
Platform
Python app (the Copilot); the full platform is in development
occupations
927
skills
258
O*NET records
96,000+

In short

Who it’s for
Someone tired and stuck in their current work who can’t see what else they could realistically do.
What was hard
Job boards show listings, not paths.
What I decided
Two parts in order (where could I go, then how do I get there), and it never applies on anyone’s behalf.

Most people don’t hate work. They’ve outgrown it.

There are millions of people who go to work every day feeling tired, stuck, and quietly unhappy. Not because they’re lazy, and not because they lack ability. They’ve simply outgrown the work they’re in, and they can’t see a way out.

When they look for help, what they find is job boards: endless lists of openings, most of them variations of the job they already want to leave. But stuck people aren’t short of listings. They’re short of a believable next step. What else could someone like me actually do? What would it take to get there?

That’s the question I’m building for.

The platform works as one journey in two parts.

First: where could I go? It reads the skills you already have and maps them against 927 occupations and 258 skills, built from more than 96,000 records of real job data. Then it shows realistic roles you could move into, and exactly which gaps you’d need to close. Not fantasy careers. Real paths that start from who you are today.

Then: how do I get there? It finds openings that genuinely fit, scores each one honestly, flags what’s missing, and helps you prepare a strong application for each. It will never apply on your behalf. Mass applications are a big part of what makes job hunting feel hopeless, for candidates and companies alike. This platform is built for fewer, better choices.

I’ve designed it to keep working even when things go wrong. If one AI service fails, it falls back to the next, and then to a simple built-in template, so nobody gets stranded halfway through a decision about their life.

Looking back, I realise this isn’t a new mission. SkillItUp was about helping skilled workers prove what they can do and find work that values it. This platform is about helping people who feel stuck find work that fits who they’ve become. Years apart, it’s the same idea: helping people get from where they are to work that’s right for them.

That’s where I’m going.

“Stuck people aren’t short of listings. They’re short of a believable next step.”

How it works

The first half, the Career Transition Copilot, is built in Python: an ETL pipeline that turned 96,000+ O*NET records into a clean knowledge base, sentence-transformer embeddings to match skills to occupations, and a shortlisting step that cut recommendation cost about 77×. The second half grows out of an AI-agent job-search workflow I’ve been using myself.

  • Python
  • sentence-transformers
  • Pydantic v2
  • spaCy
  • SQLite
  • Streamlit
  • Plotly

The proof

occupations
927
skills
258
O*NET records
96,000+

The test suite grew from 45 to 100, catching bugs like a pattern that mistook the word “me” for a Master’s degree.

What I learned

SkillItUp and this are the same mission years apart: helping people reach work that’s right for them.

An idea that won’t leave you alone?