Explore live interactive architecture drawers, verified certifications, and explainable AI pipelines:
svkhun.vercel.app

"I am a Computer Engineering student at Srinakharinwirot University specializing in Data Science, Machine Learning, and Data Engineering. I focus on developing production-grade pipelines, sub-80ms real-time Graph Neural Networks (RGCN), dynamic credit & reliability risk assessment systems, and transparent, explainable AI architectures tailored for the financial services and e-commerce logistics industries."

• Education: B.Eng. in Computer Engineering, Faculty of Engineering, Srinakharinwirot University (SWU) (2022 – Present).
• Target Roles: Aspiring Data Scientist / Machine Learning Engineer seeking technical opportunities and internships within Banking, FinTech, E-Commerce, and Risk Analytics.
• Core Domains: Relational Graph Intelligence (RGCN / PyG), High-Throughput Sub-80ms Inference (ONNX Runtime / Triton), Credit & Reliability Risk Modeling (PD, COD failure forecasting), Explainable AI (SHAP, LIME), and Production Data Engineering.
• Problem Solving: Multi-competition hackathon finalist experienced in optimizing tabular & graph prediction models under strict evaluation metrics (AUC) and production latency SLAs (P99 < 12ms).
• Languages: Thai (Native), English (Professional Working Proficiency, CEFR B1+ / Oxford Test Score: 51), and Japanese (Beginner).

• Shopee Ultimate Case Challenge 2026 (SUCC) -- Lead Data Scientist & Product Architect (Sep 2026)
  (Live Platform: Shopee Smart COD Sentinel)
  • Architected the Smart COD Reliability System & Risk Engine to solve Cash on Delivery (COD) delivery failure disparity (10.6× higher failure risk than prepaid, 85.1% concentration of platform return-to-origin waste).
  • Machine Learning Risk Engine: Developed a calibrated Random Forest classification pipeline (ROC-AUC 0.8629, PR-AUC 0.3722, Brier Score 0.021) that converts delivery failure probabilities into an explainable Reliability Score (0–100).
  • 4-Tier Progressive Policy & Feedback Loop: Engineered a dynamic decision framework categorizing buyers into 4 operational tiers (Grade A Zero Friction to Grade D ฿40 Deposit / Prepaid Conversion) paired with an automated self-correcting feedback loop (+8 pts on delivery, -20 pts on RTO).
  • Interactive Mobile App Simulator & Production Deployment: Built a full-stack executive presentation and live mobile checkout simulator deployed on Render with FastAPI REST endpoints; validated empirical hypothesis via two-proportion Z-test ($z = 24.42, p &lt; 0.001$), projecting a 58% RTO waste reduction and ฿42.9M annual EBITDA boost.
• KBTG Kampus Hackathon 2026 -- Track 2: Data Science & Intelligence (Sep 2026)
  • Architected FlowSense & TrustGraph for K PLUS, an enterprise-grade digital banking platform eliminating Budget Burnout and Security Friction for First Jobbers (18k–35k THB).
  • Module A (FlowSense): Engineered 30-day liquidity forecasting via LightGBM with rolling-window lag features & transaction seasonality; created the Status Horizon Bar, suppressed alert fatigue during normal dips, and deployed automated micro-sweeps with 1-Tap Undo (100% instant recall, zero penalties) into a high-interest sub-account (1.50% p.a.).
  • Module B (TrustGraph): Developed a Relational Graph Convolutional Network (R-GCN via PyG) to detect multi-tier mule account topology and velocity in real-time; achieved 3.85ms median (P99 11.62ms) inference latency on Triton/ONNX serving, beating the strict <80ms core banking SLA.
  • Zero-Delay & User Autonomy: Built Zero-Delay Baseline for routine transfers (0 added steps) and replaced arbitrary 15-minute lockouts with Micro-Auth (5s face liveness) and transparent Direct Risk Reasoning in plain language, preserving user autonomy while stopping task scams.
  • Deployed full-stack production architecture with Kafka event streaming, Feast Feature Store (Redis O(1)), FastAPI, and an editorial Bento Grid React 18 SPA on Render.
• Aihack Thailand 2025 (Final Round) -- Organized by AIRA & AIFUL, Chulalongkorn Business School (Dec 2025)
  • Advanced through the Selection Round to compete as a national top 8 finalist in an intensive 3-day data science competition.
  • Developed machine learning classification models on real-world financial loan datasets to predict Long Overdue Debtors (LOD, >90 days overdue after 12 months).
  • Engineered predictive features, optimized pipeline performance for AUC leaderboard rankings across Public and Private test datasets on Probspace, and delivered strategic business insights.
• True Innovation Launchpad 2026 -- AI & Data Science Lead (Team Nakphatthana Tuapralat) (Jan 2026)
  • Spearheaded data strategy and predictive modeling for elderly frailty prevention in the Healthcare Innovation Track, developing sub-30ms clinical risk scoring APIs.
• Geospatial Intelligence for Resilience Hackathon 2026 -- Spatial Data Scientist & ML Modeler (Jan 2026)
  • Organized by GISTDA x KMITL x KMUTT; built "Rain-to-Flood" spatial model fusing Sentinel-1 SAR soil backscatter, Copernicus GLO-30 DEM, and precipitation APIs for 30m flash flood mapping.
• CDG Hackathon 2026 -- Team GrandGuardianAI (Jul 2026)
  • Competed in architecting specialized AI solutions targeting high-impact public sector and operational challenges.
• LINE MAN Wongnai Junior Case Competition -- Strategic Growth & Unit Economics Lead (Team Low Cortisol) (Nov 2024)
  • Formulated campus delivery optimization and "Silent Mission" growth campaign with 8.87 THB/user CAC, acquiring 462 VIP trial conversions.