Get Started

Kia’s Fleet Management System

Problem Space

Kia’s legacy fleet system struggled with:

  • Static routing (ignoring real-time traffic/weather)
  • Reactive maintenance (costing $4,200/vehicle/year in downtime)
  • Driver inefficiencies (15% excess fuel burn from poor behavior tracking)

As the AI Team Leader Consultant for Kia’s next-generation fleet management platform, I spearheaded a cross-functional effort to transform raw telematics data into an intelligent, generative AI-powered command center. Here’s how we did it:

The Challenge

When I took over Kia’s fleet AI division, we faced three critical gaps:

  • Static Operations: Routes were planned weekly, blind to real-time disruptions
  • Costly Breakdowns: 27% of maintenance was reactive, averaging $4,200 per incident
  • Driver Blindspots: No systematic way to improve fuel efficiency or safety

My team’s mandate: Build an AI brain that could see, predict, and guide fleet operations autonomously.

Key Technical Decisions I Championed

1. Hybrid AI Architecture

  • Why: Needed both precise numerical predictions and human-readable insights
  • How:
    • Used Prophet for time-series forecasting of vehicle health
    • Built fine-tuned LLaMA-2 models to explain predictions to dispatchers
    • Developed custom APIs to connect with Kia’s legacy systems

2. Edge AI Deployment

  • Challenge: Most fleets lacked constant cloud connectivity
  • Solution:
    • Quantized models to run on NVIDIA Jetson modules in vehicles
    • Designed delta updates to sync data when connectivity available

3. Responsible AI Framework

  • Implemented:
    • Bias audits for driver scoring algorithms
    • Explainability reports for EU AI Act compliance
    • Anonymization pipelines for driver privacy

Leading Through Challenges

1. Data Silos Breakdown

  • Action: Personally negotiated with 5 department heads to unify data access
  • Outcome: Created Kia’s first federated learning pipeline across divisions

2. Model Drift Crisis (Month 9)

  • Situation: Routing accuracy dropped 15% after a telematics update
  • Response:
    • Instituted daily monitoring dashboards
    • Deployed automated retraining triggers
    • Reduced drift incidents by 80%

3. Driver Adoption Hurdles

  • Innovation: Developed AI “co-pilot” voice assistant that:
    • Provided real-time feedback in conversational Korean/English
    • Included gamified leaderboards to incentivize improvement

Results We Delivered Together

  • $28M Annual Savings (Validated by Kia Finance)
  • 92% Fleet Uptime (From 73% baseline)
  • 17% Safer Driving Scores Across 12,000 Vehicles
  • White-Labeled for Kia’s Global B2B Partners