Context
This client operates a chain of gaming lounges in airports and wanted to streamline the process of applying targeted coupons and perks based on real passenger activity.
Issue:
Manual validation of boarding passes was slow, error-prone, and limited in its ability to personalize offers.
Approach
Applied AI for Real-World Problem:
- Scoped an MVP for automating boarding pass interpretation using OpenAI’s natural language and vision models.
- Collaborated with company team to map the end-to-end reward workflow and identify security and regulatory constraints.
Integration Strategy:
- Designed MVP to scan and read with OpenAI API for real-time boarding pass parsing, validation, and account setup.
- Defined logic for eligibility checks and coupon application based on extracted boarding pass data.
Iterative Experimentation:
- Ran controlled pilots to benchmark accuracy, speed, and user acceptance against manual methods.
- Incorporated feedback from stakeholders to refine the model and UI/UX.
Execution
- Led integration of OpenAI API for text and structured data extraction from boarding pass images (PDF and JPEG).
- Built middleware to process outputs, check eligibility, and trigger targeted coupon assignments in customer accounts.
- Implemented robust error handling and fallback manual review for edge cases.
- Coordinated with compliance/security to ensure full GDPR and aviation industry alignment.
Outcome
- Reduced average coupon processing time from 3 minutes to under 30 seconds.
- Improved accuracy of eligibility validation to 98%+, minimizing false positives/negatives.
- Increased customer coupon redemption rates by 24% via more timely and personalized offers.
- Freed up staff resources for higher-value customer service tasks.
Tools & Frameworks Used
- OpenAI API (Vision, NLP)
- Python/Node.js (integration & backend processing)
- REST API, secure middleware
- Google Cloud (deployment & logging)
Capabilities Demonstrated
- Primary: Artificial Intelligence (AI)
- Secondary: NPD, Optimization, Product Design
Key Learnings
- Real-world AI productization depends on clear scoping, stakeholder buy-in, and iterative model refinement.
- User feedback and exception handling are as critical as raw model accuracy for operational AI adoption.
- Security and compliance must be integrated from day one—especially in regulated environments like airports.