Extract the profile
PDF/DOCX resume data becomes structured skills and experience.
IntelliApply
FULL STACK / RELEVANCE RANKINGLive application
From who you are to what comes next.
THE CONTEXT
Help a candidate turn collected job postings into a relevant shortlist.
THE KEY DECISION
Rank textual relevance with TF-IDF and cosine similarity; retain user decisions.
THE RESULT
An authenticated resume-to-matches workflow, verified end to end by the project owner.
IntelliApplyPROFILE → RELEVANCE
MECHANISM STUDY / 06SCHEMATIC
TAKE A CLOSER LOOKSELECT A STAGE ↓
SCROLL TO TRACE. SELECT TO EXPLORE.
Follow the connections as you scroll, or select a stage to see what happens there.
Compare the openings
TF-IDF and cosine similarity compare the candidate with collected jobs.
Keep the shortlist
Ranked results and interested/applied/ignored states live in the dashboard.
A working full-stack job discovery application connecting resume parsing, job collection, relevance ranking and personal job-state tracking.
PROJECT TYPE
Full-stack product engineering
FOCUS
Full stack + Backend + AI / ML
STATUS
Live application
/01
The question.
A growing list of openings is not a useful shortlist. The application needs to compare a candidate’s skills and experience with collected postings, then retain that candidate’s decisions over time.
/02
The approach.
A React/Vite interface uses Neon Auth with a FastAPI backend and PostgreSQL. Resume text feeds the candidate profile; job sources feed normalized openings. TF-IDF vectors and cosine similarity rank the matches, while background refresh and matching tasks keep the dashboard current.
/03
What came out of it.
The authenticated application connects resume context to ranked job results, match counts and refresh workflows. Its end-to-end flow has been verified by the project owner. Match scores describe textual relevance and remain distinct from hiring outcomes.
01
A concrete matching method
The matching implementation uses TF-IDF and cosine similarity. The illustration shows relative ordering, with no invented percentages or accuracy claims.
02
Identity reaches the API
Neon Auth/Better Auth sessions are verified by the backend before accessing candidate-specific records.
03
Discovery is a background workflow
Collection, refresh and matching support the product beyond the initial upload. Job tracking preserves the user’s decisions.
Under the surface.
- Experience
- React · Vite · TypeScript
- Identity & persistence
- Neon Auth / Better Auth · PostgreSQL
- Matching & collection
- FastAPI · resume parsing · scikit-learn · scheduled tasks
SCOPE & LIMITS
- Ranking is a relevance heuristic, not a guarantee of hiring outcomes.
- Availability of collected postings depends on external sources.
- Authenticated end-to-end verification is user-reported; the portfolio does not invent dashboard metrics.
NEXT PROJECT / 07
Have an idea of your own? Let’s talk