About Find your match
An AI-powered job aggregator for fresh graduates in Bangladesh.
The problem
BD job boards list thousands of openings across dozens of categories. Most are wrong for any individual candidate — wrong field, wrong experience level, wrong role type. Scrolling through pages of irrelevant listings to find the few that fit is genuinely the worst part of job hunting.
How it works
1. Scrape the boards
Every 12 hours, an automated scraper pulls the latest listings from Bdjobs, Skill.jobs, and Bdjobstoday across IT, healthcare, engineering, marketing, and other categories. Each job is stored with its real description.
2. Read your CV
When you upload a CV, GPT-4o-mini reads it and extracts a structured profile: your career level, target job categories, technical keywords, and role types to avoid (e.g. a SOC analyst's profile is told to avoid "sales", "teaching", etc.). The raw CV is never stored.
3. Vector + keyword search
Both your CV and every job become a 1,536-dimensional vector (an embedding). We find the 60 jobs whose vector is closest to yours, then rerank by exact keyword overlap, fresh-grad-friendly titles, and required experience years.
4. AI judges role fit
The top candidates go through one more LLM pass that grades each job on role fit (not just keyword overlap). It catches mistakes like "Sales Engineer - Cybersecurity" looking relevant by keywords but being the wrong role family. The final score is a blend of the algorithmic and AI scores.
Privacy
Your CV file and its raw text are never stored on our servers — they're sent to OpenAI to extract a profile and then discarded. Cover-letter drafting briefly keeps the CV text in your browser's session storage so you don't have to re-upload between matches; it's cleared when you close the tab or when you click "Clear my CV".
If you create an account and choose to save a search, we store the structured analysis (career level, target categories, keywords) plus the IDs of the 20 matched jobs — never the raw CV text. If you star a job while signed in, we store a copy of that job listing in your account so you can find it again later. You can delete any saved search or starred job at any time. Signed-out stars live only in your browser.
Stack
Next.js, Supabase (Postgres + pgvector + Auth), OpenAI (embeddings + gpt-4o-mini), Playwright for some scrapers, plain fetch for others. Hosted on Vercel. Scrapers run as scheduled GitHub Actions.
Who built this
A recent CSE graduate in Bangladesh, built as a personal project to learn modern full-stack and AI tooling — and out of frustration with how BD job boards work for fresh grads.