CareerCraft: Supporting New Graduates on Job Hunting with LLM-Assisted Self-Construction of Career Profile

Paper Title

CareerCraft: Supporting New Graduates on Job Hunting with LLM-Assisted Self-Construction of Career Profile

Publication Info

• Topic area: Career profile construction and job hunting support for new graduates using LLMs.
• Keywords: Career profiles, job hunting, large language models, self-exploration, experience cards, career readiness, user agency, reflective systems, human-computer interaction, AI-assisted tools.

Background and Problem

• Problem / challenge: New graduates face significant challenges in constructing actionable career profiles due to limited self-awareness, fragmented experiences, and difficulty aligning their skills with job requirements. Existing tools, such as LinkedIn, are more suited for experienced professionals and fail to address the unique needs of early-career individuals.
• Significance: Helping new graduates construct coherent career profiles is critical for improving their job readiness and enabling informed career decisions. This issue is particularly important as it impacts their ability to transition effectively from education to the workforce.
• Motivation and related work: Prior research and tools like CareerSim, CareerBERT, and JobFit have explored career guidance and semantic matching but lack interpretability and user agency. Current solutions do not adequately support new graduates in integrating diverse, unstructured experiences into meaningful career profiles. This paper addresses these gaps by introducing CareerCraft.

Solution

• Proposed approach: CareerCraft, an LLM-powered system that assists new graduates in constructing structured career profiles through experience card extraction, guided narrative building, and personalized job recommendations.
• Novelty:
  1. Development of a multi-stage pipeline to transform fragmented experiences into structured career profiles.
  2. Introduction of experience cards to capture formal and informal experiences, enabling reflective and iterative profile construction.
  3. Integration of dynamic job recommendations tailored to user profiles and career goals.
• Procedure and key techniques:
  • Stage 1: Users input text, files, or images to extract experiences into structured "experience cards" using LLMs.
  • Stage 2: Experience cards are grouped, refined, and combined into coherent career narratives.
  • Stage 3: Career profiles are visualized with radar charts and quadrant diagrams, and linked to personalized job recommendations with match scores and rationale.

Results

• Concrete findings:
  • CareerCraft improved relevance judgment (M**CareerCraft = 5.81, Mbaseline = 4.49, p <.01), hidden insights (M**CareerCraft = 5.94, Mbaseline = 4.25, p <.01), and reflective depth (M**CareerCraft = 6.31, Mbaseline = 4.25, p <.01).
  • Enhanced career clarity (M**CareerCraft = 6.00, Mbaseline = 4.62, p <.01) and readiness confidence (M**CareerCraft = 5.94, Mbaseline = 3.88, p <.01).
• Advantage over baselines:
  • CareerCraft outperformed traditional methods in experience structuring (+1.31 higher), decision grounding (+1.93 higher), and match satisfaction (+1.81 higher).
  • Participants reported broader career exploration, deeper reflection, and more coherent profiles compared to baseline practices.
• Experiments / evaluation:
  • Conducted a within-subject user study with 16 participants (recent graduates without formal work experience).
  • Participants completed career profile tasks in two conditions (baseline and CareerCraft) and were evaluated using Likert-scale questionnaires and semi-structured interviews.
  • Metrics included clarity of experience sorting, confidence in material selection, and perceived career readiness.
• Limitations and future work:
  • Focused on short-term self-exploration; long-term impacts on career development remain unexplored.
  • Limited to recent bachelor’s graduates; future work should include diverse user populations and address AI literacy barriers.
  • Integration with real-world job platforms and exploration of competency-based social proof are suggested for future iterations.

Summary

CareerCraft is an LLM-powered system designed to help new graduates construct actionable career profiles by transforming fragmented experiences into structured narratives and providing tailored job recommendations. The system demonstrated significant improvements in self-exploration, experience structuring, and career clarity compared to baseline methods. Through a mixed-methods evaluation, participants reported increased confidence, broader career exploration, and higher-quality career decisions. CareerCraft’s design principles, emphasizing user agency, transparency, and adaptive scaffolding, offer valuable insights for developing future self-exploration tools. While effective in the short term, future research should address long-term impacts, inclusivity, and integration with real-world job platforms.