AI Engineer with experience applying machine learning to solve lending and credit risk problems. Analytical and self-motivated, with experience building ML pipelines, developing and comparing models to support decision-making, and working on Deep Learning and Generative AI , with additional background in OCR/image processing and API-based internal tools.
Education
Chulalongkorn UniversityExpected May 2026
B.Sc. in Computer Science
Work Experience
AI Engineer Intern | ClickNext Co., Ltd.Jan 2026 - Present
OCR & Frontend Services: Benchmarked and fine-tuned a Thai handwriting OCR model, reducing the Character Error Rate from 60.95% to 19.98%, and successfully deployed the solution through a Next.js frontend service.
Computer Vision & Speech Processing: Developed a computer vision and speech processing pipeline to extract structured data from live video clips.
Backend Development: Built a meeting summarization service utilizing FastAPI and LLM APIs, implementing structured text processing pipelines to optimize internal workflows.
Projects
Credit Risk Assessment & Lending SolutionsJan 2026 - Feb 2026
Built a predictive credit risk pipeline using machine learning models to generate reliable Probability of Default (PD) assessments from large financial datasets.
Enhanced lending decision strategies by integrating threshold analysis with risk-adjusted return optimization, enabling data-driven loan approvals that mitigate complex financial risks.
Applied isotonic regression calibration to ensure reliable and interpretable risk scores.
Identified an optimal PD threshold (22%) to maximize risk-adjusted returns and expected loan collection.
Video Anomaly Detection System using SlowFast and Adaptive MissionGNNAug 2025 - Mar 2026
Built a feature extraction pipeline using SlowFast (temporal) and ImageBind (semantics), generating high-dimensional combined representations from surveillance video data.
Extended the MissionGNN research model by concatenating SlowFast features with ImageBind features and evaluating the impact on anomaly detection.
Achieved 0.8484 mAUC on UCF-Crime and 0.7367 mAP on XD-Violence datasets.
Multi-Label Emotion Detection in Tweets (BERTweetMA)Jan 2025 - Feb 2025
Built a multi-label classification model using BERTweet with attention mechanisms on 38K tweets.
Designed emotion-specific attention blocks with sigmoid-based classification.