Job Atlas
A job hunt dashboard that reads your inbox for you. Applications, replies and rejections tracked without filing anything by hand.
Product
I built this because my own job search was a spreadsheet I stopped updating in week two. tracking applications isn't hard, it's just a second job you end up doing badly while doing the first one badly.
so Job Atlas does the filing. it connects to Gmail over OAuth, scans the inbox on a schedule, and sorts what it finds (confirmation, interview invite, rejection, offer) into a timeline per company. you apply, the tracker catches up on its own.
around that: a cold-email engine with template tags that fill from a parsed résumé, a résumé manager, a contacts book, and an extractor that turns a pasted job listing into structured rows you can save.
Next.js and shadcn on the front, FastAPI and SQLAlchemy behind it, Postgres with pgvector for matching, and Redis with Celery running the two hourly inbox sync and the dispatch queue. the model provider is a config value, Gemini or OpenAI, because I didn't want the product's behaviour welded to one vendor's pricing page.
What I’d point at
the template tags. they fill from the parsed résumé with plain string replacement, no model call per send. paying an LLM to do find-and-replace is the easiest money to waste in a product like this.