Tech Blog
Vitalify Asia Tech Blog
Engineering notes, product thinking and offshore development insights from the Vitalify Asia team.
82 articles
Latest articles
3DWhat Is 3D Tiles? A Complete Guide to Its Core Concepts, Latest Trends, and Use Cases
3D Tiles is an open standard for streaming massive 3D geospatial data. This guide explains HLOD, bounding volumes, tileset.json, differences between versions 1.0 and 1.1, use cases including PLATEAU and Google Photorealistic 3D Tiles, and independent LOD control in CesiumJS.
AI DevelopmentWhat We Learned Building Voice AI with Gemini: The Major Difference Between a PoC and a Commercial Service
A Gemini Live API PoC can deliver real-time voice conversations quickly, but production introduces concurrency, quotas, observability, model lifecycle and device-specific audio issues. This article explains why the architecture moved from direct browser access to LiveKit and Vertex AI.
Business & CultureWhat I Did to Pass the PMP Exam
Nihei, a project manager at Vitalify Asia, explains how he prepared for and passed the PMP exam. From the 35-hour course and a Japanese exam guide to repeated Study Hall practice and taking the test in Bangkok, this article covers the full journey and lessons learned.
AI DevelopmentA Concept for Developing AI Through Artificial Languages
Can AI learn logic more efficiently through an artificial language than through natural language? This article compares Esperanto, Lojban, and Ithkuil, then presents a custom GPT-2 model trained from scratch on Lojban-based data that achieved 100% accuracy on prepared three-valued logic tests.
AI Development7 AI Image-to-3D Generators I Tested in 2026: Which One Is Actually Worth Using?
Using the same portrait and default settings, we tested Prism 3.1, Meshy 6, Tripo P1, Hunyuan3D, Forge, Trellis 2 and Rodin 2.5. See how they compare in facial similarity, texture, hair, clothing and mesh quality, plus which tools work best for realism, game assets and further editing.
Product DevelopmentGLB Texture Editor: Edit 3D Model Textures and Materials in Real Time in Your Browser
GLB Texture Editor is a browser-based tool for editing 3D model textures and materials in real time. Upload a GLB file, adjust base color, roughness, metalness, transparency, UV mapping, position, rotation, scale, and projection settings, then preview and export the updated model.
Digital TwinHigh-Performance Web BIM Viewer for Large IFC Models
This PoC converts large IFC models into optimized GLB files by BIM category for faster browser rendering. It demonstrates flexible layer controls, element data panels, smooth 3D navigation, and reduced z-fighting, while outlining a future automated IFC-to-GLB workflow.
Business & CultureSPORTEC Thailand 2026: ASEAN Sport Tech Report
A field report from SPORTEC Thailand 2026 in Bangkok. We explore how ASEAN’s aging market is shifting Sport Tech from equipment to software and data, and how Taiwanese companies and Japan’s wellness expertise are shaping new business opportunities.
AI DevelopmentConstellation Recognition: GNN + Symbolic AI — Developing a Hybrid Model
We developed a hybrid AI system that detects, separates, and reconstructs Orion, Cassiopeia, and the Big Dipper from noisy, rotated, and scaled point clouds. It combines geometric signatures, an edge-updating GNN, and symbolic graph-isomorphism constraints, reaching a test macro F1 score of 96.91%.
AI DevelopmentWhat Is Function Calling? How AI Agents Work and How to Implement Them with LangGraph
Learn how Function Calling lets LLMs use external tools instead of performing every task themselves. This guide covers tool-use architecture, the evolution of coding agents, Structured Outputs, and a practical LangGraph and TypeScript implementation of the reasoning-and-action loop.
AI DevelopmentAutomatically Generate 3D Models by Inferring Noisy Floor Plans! A Challenge in CubiCasa5K Analysis and Reverse Engineering
Using only floor plan images and SVG annotations—and without reading the official paper or code—the author reverse-engineered CubiCasa5K. The article covers semantic segmentation, random-rotation augmentation, inference of an incomplete sauna door, and conversion of the final result into a 3D model.
AI Development3D Avatar from One Photo with FLAME, DECA and DiffLocks
From a single face photo, we tested a workflow for creating a realistic 3D head and natural-looking hair using FLAME, DECA and DiffLocks. This article covers the process, results, texture issues, hair generation challenges and license notes.
