Shiyong Z.

Key Skills

Software

Other

Top Subject Matter

Affective computing

emotion recognition from EEG and speech

AI-assisted security compliance evaluation

Top Data Types

Video

Text

Document

Top Task Types

Emotion Recognition

Freelancer Overview

NTU Final Year Project — EEG-Driven Continuous Emotion Prediction During video viewing (CBCR lab). Brings 8+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other. Education includes Bachelor of Computing (Honors), Nanyang Technological University (2025) and Diploma in Information Technology, Singapore Polytechnic (2020). AI-training focus includes data types such as Medical, DICOM, and Computer Code and labeling workflows including Emotion Recognition, Evaluation, and Rating.

Labeling Experience

Cyber Security Engineer — AI-assisted automation PoC for STIG compliance comparison/testing

Other

Conducted a proof-of-concept for AI-assisted automation to evaluate feasibility of automating security compliance comparisons and health checks. Designed and executed experiments in a testing environment using an AI coding workflow to support STIG compliance evaluation tasks. Integrated Playwright-based web automation to run and validate the proposed compliance checking approach. • Focused on evaluating outputs of automated security health checks for compliance • Used AI-assisted automation (Claude Code) to support workflow implementation • Applied web automation to collect/verify comparison results in tests • Collaborated on validation and troubleshooting within security monitoring workflows

2025 - 2026

NTU Final Year Project — EEG-Driven Continuous Emotion Prediction During video viewing (CBCR lab)

OtherEmotion Recognition

Built an adaptive, real-time emotion prediction system using multimodal EEG and synchronized speech analysis. Developed deep learning models to interpret and annotate complex emotional states for continuous prediction during video viewing. Evaluated predictive performance to achieve 78% classification accuracy. • Data included continuous EEG recordings and speech-derived prosodic/semantic cues • Target labels represented discrete/continuous emotion states inferred from multimodal inputs • Used deep learning to map signals to emotion annotations in real time • Validated feasibility and effectiveness via classification accuracy results

2024 - 2025

Education

Work History