CareerEngine

Status: In active development — V2.1, June 2026

CareerEngine is a job-discovery and ranking pipeline built for the developer to help with early-career software engineering search.

It collects active roles from validated public job sources, evaluates them against a candidate profile, ranks them, and generates a daily opportunity report. The goal is to reduce the time spent manually checking careers pages while keeping application decisions fully manual.

What it does

• Collects active roles from public ATS platforms including Greenhouse, Ashby, Lever, and Workable.
• Normalizes job listings into a shared format.
• Filters for software engineering, backend, full-stack, data, AI/ML, developer tools, infrastructure, healthtech, and fintech-related roles.
• Scores roles using title fit, technical overlap, degree relevance, seniority signals, work-authorization wording, and profile-to-description similarity.
• Prioritizes explicit internships, new-grad roles, SWE I, and other early-career opportunities.
• Prevents senior, staff, principal, lead, manager, and director roles from receiving early-career bonuses.
• Sends the top-ranked opportunities in a daily email report.
• Generates a ranked TXT attachment for additional qualified opportunities below the email cutoff.
• Tracks jobs shown in the main email body to prevent repeat recommendations.
• Records source health, delivery state, and recommendation diagnostics.

Ranking flow

CareerEngine builds one shared candidate pool each day:

Active qualified jobs
- roles already shown in an email body
= ranked candidate pool

The ranked list is split into:

#1–#25   Daily email body
#26–#N   Additional qualified opportunities attachment

Roles in the attachment remain eligible for future runs. They are ranked again alongside newly discovered jobs rather than being treated as permanently delivered.

Ranking signals

The current scoring model considers:

• Software engineering and backend relevance
• Python, Java, TypeScript, SQL, REST APIs, PostgreSQL, Docker, cloud systems, data pipelines, and ML/LLM overlap
• Computer Science and Mathematics degree relevance
• Internship and early-career signals
• Seniority penalties and blockers
• Work-authorization wording when present
• Lexical profile-to-description similarity

Sentence-BERT semantic similarity is currently running in shadow mode. It is calculated for evaluation but does not yet affect the live recommendation order.

Tech stack

• Python
• SQLite
• GitHub Actions
• Greenhouse, Ashby, Lever, and Workable public job sources
• Sentence-Transformers with all-MiniLM-L6-v2
• pytest
• SMTP email delivery

Local setup

git clone https://github.com/cerenoguz/careerengine.git
cd careerengine

python -m venv .venv
source .venv/bin/activate

pip install -r requirements.txt

cp config/candidate_profile.example.txt config/candidate_profile.txt

python -m pytest
python -m src.main

Local runs generate reports without sending email unless SMTP environment variables are configured.

Project structure

config/
  companies.yaml
  candidate_profile.txt

src/
  collectors/      Public ATS collectors
  ranking/         Fit scoring and semantic similarity
  reporting/       Email and attachment generation
  storage/         SQLite delivery and deduplication state
  compliance/      Source-access safeguards

tests/
  Collector, ranking, reporting, delivery, and backlog tests

scripts/
  Source validation and maintenance utilities

Compliance

CareerEngine only uses validated public job sources. It does not access logged-in pages, bypass CAPTCHAs, rotate proxies, evade rate limits, or scrape blocked sources.

See LEGAL_AND_COMPLIANCE.md, TERMS_OF_USE.md, and robots_policy.md for project policies.

Current status

Implemented:

• Public ATS job collection
• Rule-based fit scoring
• Early-career ranking calibration
• Daily email reporting
• Ranked additional-opportunities attachment
• Seen-job tracking and delivery auditing
• Sentence-BERT semantic matching in shadow mode
• GitHub Actions automation
• Test coverage for ranking, delivery, reports, and collectors

In progress:

• Production validation of the revised backlog behavior
• Controlled rollout of semantic ranking
• Compliant collectors for large companies with custom careers sites

Notes for contributors

config/candidate_profile.txt, local SQLite files, generated reports, and SMTP credentials are local runtime configuration. They should not be committed to a public repository.