karthik.dev
Back to projects
Open Source

Career Agents

Open-source multi-agent system automating the job search pipeline

GitHub
Multi-Agent SystemsOpenAI APINode.jsTypeScriptREST APIs

Overview

Career Agents coordinates a set of specialized AI agents to take a candidate from raw profile data to a fully prepared job search — resume tailoring, interview simulation, LinkedIn optimization, and application tracking — with each agent owning one stage of the pipeline.

Problem

Job seekers juggle disconnected tools for resume writing, interview prep, LinkedIn optimization, and tracking applications, redoing the same context-gathering work for every tool.

Solution

A pipeline of cooperating agents — Resume Agent, Interview Agent, LinkedIn Agent, and Job Tracker — that share a single candidate context object, each specializing in one transformation and handing structured output to the next agent in the chain.

Architecture

  1. 1User submits profile + target role
  2. 2Career Agent (orchestrator) decomposes the request and dispatches to specialist agents
  3. 3Resume Agent tailors resume bullets and keywords to the target job description
  4. 4Interview Agent generates likely questions and model answers based on the role and resume
  5. 5LinkedIn Agent proposes headline/summary optimizations for discoverability
  6. 6Job Tracker persists application state and follow-up reminders

Technical Challenges

  • Keeping agent handoffs deterministic — passing structured JSON contracts between agents instead of free-form text to avoid context drift.
  • Controlling latency and cost across multiple chained LLM calls by batching and caching intermediate agent outputs.
  • Designing prompts that stay consistent across very different candidate backgrounds without hand-tuning per user.

Results

  • Open-sourced on GitHub for the developer community to extend with new agent types.
  • Demonstrates a reusable orchestrator + specialist-agent pattern applicable beyond job search.