Nissorn Koraviyothin

AI Model Developer

Singapodent Digital Internship · Vietnam · April 2026 – June 2026

Python · PyTorch · nnU-Net · HRNet · YOLOv8 · Computer Vision

• Built an end-to-end 2D cephalometry AI pipeline (PoC) with a cascaded architecture (YOLOv8, HRNet, nnU-Net) for landmark detection and anatomical segmentation.
• Mitigated domain shift and scanner variability in dental X-ray datasets through targeted augmentation.
• Implemented geometric algorithms for alveolar bone thickness measurement plus a Human-in-the-Loop fallback UI for out-of-distribution images.

Cognitive Fingerprint: Continuous Stylometric Authentication

2026

Python · PyTorch · XGBoost · Char-CNN · Siamese Network

• Engineered a heterogeneous late fusion system (character-level CNN, TF-IDF stacking, and behavioral heuristics) to extract Thai cognitive fingerprints for continuous authentication.
• Trained a Siamese network with online hard negative mining and k-NN clustering to separate genuine users from impostors.
• Reached 97.70% verification accuracy with a 2.24% false acceptance rate, using Llama 3.1 data augmentation and time-decay weighting to resolve cold start within 10-15 messages.

GrowthSheet: AI-Powered Study Sheets Marketplace

2026

Python · Typhoon LLM · OCR · ASR · Microservices

• Built a microservices AI backend with RESTful APIs for automated Thai OCR and ASR content extraction.
• Integrated the Typhoon LLM for summarization with context-aware memory for continuous chatbot interactions.

AI-Powered C++ Plagiarism Detection

2025

PyTorch · Transformers · Scikit-learn

• Trained a Siamese CodeBERT model with mean pooling to capture deep semantic similarities in C++ source code, detecting paraphrased plagiarism beyond token matching.
• Reached 94.03% F1 and 93.93% accuracy, outperforming the Scikit-learn baseline.
• Parsed program structure with Tree-sitter ASTs and tuned dynamic cosine similarity thresholds for classification precision.

AMBA: Generative AI Musical Box

2025

Python · FastAPI · Markov Chains · IoT

• Engineered an emotion-driven melody generator using a fourth-order Markov chain with dynamic mood filtering.
• Served a FastAPI harmonic memory system to maintain melodic continuity across stateless requests, streamed to an IoT musical box.

Technical Team & Data Analyst

Cybergeek Club · 2024–2025

• Instructed the Prompt Wizard workshop, teaching practical prompt engineering to club members.
• Ran 2025 technical operations for club events and competitions.
• Analyzed 2024 engagement data to shape recruitment and activity planning.

Participant

TESA Top Gun Rally · November 2025

• Gained hands-on embedded systems experience applying MATLAB to solve problems under competition time constraints.
• Advised the team on implementing the drone detection feature with a YOLO model.

Data Science Bootcamp

DataRockie School · 2026

• Mastered data manipulation and visualization with SQL, spreadsheets, and Google Looker Studio to deliver actionable business insights.
• Applied Python and R for exploratory data analysis, statistical testing, and predictive models covering classification, regression, and clustering.
• Leveraged prompt engineering and automated workflows to optimize data processing and analysis.

Data Scientist: Machine Learning (100+ Hours)

Codecademy · 2025

• Completed 100+ hours of machine learning coursework spanning supervised and unsupervised algorithms.
• Built end-to-end projects covering feature engineering, model evaluation, and reporting.