AI Career CoPilot
Multi-Agent Job Discovery & Kanban Tracking Pipeline Engine
// 01. THE PROBLEM
Applying for technical jobs requires hours of manual work tailoring cover letters, matching key experience items to job specs, and tracking application state.
// 02. SYSTEM ARCHITECTURE & PIPELINE FLOW
Job Specification Parser
Extracts core requirements, tech stacks, and seniority level from posted job URLs.
ChromaDB Vector Retrieval
Queries candidate project database for 3 most relevant experience items.
Multi-Agent Resume & Letter Generation
Orchestrates ATS parser, bullet optimizer, and letter writer agents asynchronously.
Kanban Status Synchronization
Updates application board state to READY_TO_APPLY with tailored assets.
// 03. TECHNICAL APPROACH & DECISION MATRIX
01Decoupled Celery Worker Architecture
HTTP 202 Accepted + Celery Workers + Redis
Asynchronous queue workers prevent HTTP gateway timeouts during multi-step agent execution.
Synchronous HTTP API Calls
Caused gateway timeouts when executing sequential LLM generation steps.
// 04. CORE ARCHITECTURAL LOGIC
Asynchronous Celery worker task decoupling multi-agent LLM pipeline execution.
1234567891011@celery_app.task(bind=True, max_retries=3, default_retry_delay=5) def orchestrate_job_tailoring_task(self, job_id: str, user_id: str): job_desc = db.fetch_job(job_id) user_projects = chroma_vector_store.query_relevant_projects(job_desc.text, k=3) # Trigger multi-agent pipeline asynchronously tailored_resume = agent_orchestrator.run( agents=["ats_parser", "bullet_optimizer", "cover_letter_writer"], context={"job": job_desc, "projects": user_projects} ) kanban_service.update_application_status(job_id, status="READY_TO_APPLY")
// 05. TRADE-OFFS & HONEST REFLECTION
Managing Redis infrastructure added deployment complexity, but completely eliminated UI freeze during agent generation.