Ahla Ko

Clinical Decision Support for Preventive Breast Cancer

• Designed and developed an end-to-end reinforcement learning–based clinical decision support system that personalizes cancer prevention strategies for high-risk patients (BRCA1/2, PALB2) by optimizing both intervention type (surgery vs. surveillance) and timing over a 45-year horizon, outperforming standard clinical guidelines by up to 4.33 quality-adjusted life years (QALYs).
• Performed Pareto analysis to quantify trade-offs between treatment aggressiveness and patient quality of life between using Pareto analysis, revealing that guideline-based strategies underperform on both cancer prevention and patient utility when individual preferences are not accounted for.
• Delivered model interpretability using SHAP to surface key decision drivers (menopausal status, breast density, patient preferences), making model recommendations transparent and defensible for clinical decision-making.

Metastatic Breast Cancer Treatment Optimization

• Designed and implemented a custom RL simulation environment (OpenAI Gymnasium) modeling multi-stage treatment transitions for metastatic cancer patients, optimizing the balance between overall survival and quality of life.
• Validated patient-centered thresholds for treatment escalation using Proximal Policy Optimization and executed Tornado Analysis to validate optimal policy robustness.
• Quantified optimal timing to transition patients to supportive care, quantifying the survival-vs-toxicity trade-off to support evidence-based end-of-life treatment decisions.

Medical Document Intelligence Platform

• Engineered an end-to-end OCR pipeline (YOLOv11 + Residual CRNN) for medical record digitization, achieving 99.75% accuracy (EMA) and reducing manual processing costs by over 90%.
• Established a “Human-in-the-Loop” validation system with an optimal confidence threshold, automating 91% of total data entries while maintaining a clinical-grade error rate below 0.2%.
• Implemented deterministic audit trails and rule-based consistency checks to ensure HIPAA-compliant data governance and traceability.
• See also: related publication under review (IEEE TII)

[Teaching & Mentorship]

• Served as Teaching Assistant for Applied Probability & Statistics for Engineers (INEG 2313 I & II); conducted grading and student evaluation across multiple semesters.
• Co-instructed introductory probability coursework during summer session, delivering lectures and supporting curriculum delivery.

Joined as an early engineer for “DatAmenity,” a B2B Revenue Management System (RMS) for the hospitality industry (Seed-stage startup).

• Developed backend services and REST APIs for a hotel revenue management platform supporting pricing intelligence across 20+ OTA channels using Flask, SQLAlchemy, and PostgreSQL.
• Collaborated with the Lead Developer to design relational database schemas and data models supporting hotel pricing, demand forecasting, and competitive market analysis.
• Built data-driven monitoring features including real-time pricing alerts, competitive pricing dashboards, demand visualization, and automated Excel reporting for hotel revenue optimization.

Applied mathematical optimization and quantitative analysis for national defense projects.

• Developed a facility location optimization model for military gunnery ranges, minimizing land costs while enforcing noise-impact constraints on residential areas.
• Synthesized and consolidated multi-source defense data across government agencies; prepared and presented strategic reports to senior officials (Deputy Directors, Administrative Officers) supporting ROK-U.S. cost-sharing negotiations.