About CareerFlow

What is CareerFlow?

CareerFlow is a career guidance platform for IT students and people exploring technical careers. It analyzes skills and academic background, compares supported career directions, identifies skill gaps, and generates personalized learning roadmaps.

CareerFlow is in Public Beta and is actively developed.

How it works

  1. CatBoost analyzes academic background and technical indicators to score supported career profiles.
  2. TF-IDF and cosine similarity compare your skills with each career’s skill profile.
  3. LightGBM estimates labor-market signals from historical vacancy data.
  4. A weighted ranking combines profile, skill and market signals so you can compare directions.
  5. Skill-gap analysis identifies learning steps; rule-based roadmaps connect them with courses.

CareerFlow provides exploratory career guidance. Its scores are not probabilities of career success. Market estimates use historical data and should not be interpreted as live vacancy availability.

Technology

  • Python
  • FastAPI
  • React
  • PostgreSQL
  • Supabase Auth
  • CatBoost
  • LightGBM
  • scikit-learn
  • Docker

Developer

Danial Yermekov

Developer & Maintainer

CareerFlow is actively developed and maintained by Danial Yermekov, with a focus on machine learning, practical career guidance, and accessible learning tools.

GitHubLinkedIn

Original project contribution credits are preserved in the GitHub README.

Questions or feedback: contact@careerflow.live

Feedback & IdeasWhat's NewProduct Roadmap

Privacy

Google or GitHub sign-in is handled by Supabase Auth. CareerFlow stores your Supabase user ID, display name and timestamps. Supabase may retain identity information including email under its own policies. CareerFlow does not store provider access tokens or passwords.

For signed-in users, PostgreSQL stores submitted profile details, recommendation results and AI context, skill gaps, course preferences and roadmap progress. Every private API request verifies your Supabase session and checks ownership. Guest results remain in server memory for up to two hours, are protected by a separate random credential and disappear on a server restart.

Feedback includes its type, optional rating, title, text, reproduction steps, publication consent, moderation status and time. Signed-in submissions may link to your account and your own recommendation. Only approved, consented reviews and feature suggestions appear publicly; account details and bug reports are not published. Public authors are shown as Anonymous; developer feedback is labeled. Votes store the authenticated account and feature reference. No public rating average is calculated.

CV uploads are parsed in memory to detect skills and experience signals. Raw CV files are not retained by application code. Extracted skills may be saved as part of a profile you submit. Language and theme use localStorage; Supabase persists and refreshes the sign-in session in browser storage. A guest credential uses sessionStorage.

Optional AI requests send your message, chat history and recommendation context through CareerFlow to Claude or Gemini. Voice transcription sends audio to Gemini. Your AI key stays in application memory until refresh, closing the page or removal, and is sent only for the AI request. Application code does not persist or log these keys. Provider processing and retention follow Anthropic or Google policies.

Account data is retained until you clear recommendation history or delete your account. Account deletion removes associated application data and the Supabase Auth identity when the server administrator key is configured. Failed deletion can be retried; contact contact@careerflow.live for help. Anonymous feedback is retained for product review; no automatic expiry is implemented. Hosting logs, backups, geographic processing locations and a formal retention schedule require maintainer review before wider release. No certification or independent security audit is claimed.

The optional Gemini preview sends your question, up to four recent messages and selected skills, career match, gaps, roadmap and courses to Google after your consent. It excludes account names, email, raw CV and academic profile fields. Do not include sensitive or confidential information. Google may use unpaid-tier inputs and outputs for product improvement and human review; with Cloud Billing enabled, they are not used for improvement but limited abuse-monitoring retention applies. We have not independently verified this project’s billing tier. A server-issued HttpOnly trial cookie lasts up to one year; PostgreSQL stores a hashed trial/account identifier and lifetime message count, temporary request reservations, daily global usage and short-lived hashed IP rate-limit counters. Clearing guest cookies can create a new allocation; global limits constrain provider usage. Signing in uses a separate verified-account allocation. Application code does not store preview conversations. Trial usage lasts until account deletion or maintainer cleanup; guest allocations have no automatic expiry.

Read the full Privacy Policy