←SELECTED WORKS•CASE 03•ai

AI Photo Tagger

An automated round-trip auto-tagging workflow for millions of personal NAS photos: auto-scanning the NAS, generating tags with Google Vertex AI, and writing tags back into EXIF metadata for universal photo search

SYNOLOGY NAS AUTO-TAGGERNAS ➔ VERTEX AI ➔ EXIF WRITEBACK
photo-tagger // watcher.py
Personal NAS Photo Auto-Scanning & EXIF Pipeline
Auto-scans personal photos on the NAS, calls Google Vertex AI for tag generation, and writes tags back into EXIF headers.● LIVE
✦ Round-Trip: NAS Scan ➔ Vertex AI Tagging ➔ EXIF Writeback ➔ Universal SearchHEALTHY
SYSTEM // AI_PHOTO_TAGGER

Automated NAS Photo Vision Pipeline

Autonomously inspects personal NAS archives, generates Vertex AI tags, and writes permanent EXIF IPTC.

Personal NAS Photo Auto-Scanning & EXIF Pipeline

Auto-scans personal photos on the NAS, calls Google Vertex AI for tag generation, and writes tags back into EXIF headers.

01 // PROVEN IMPACTVERIFIED
0% Embedded EXIF

Eliminated manual classification across millions of personal photos by auto-scanning the NAS, generating Vertex AI tags, and writing them straight into EXIF metadata for universal search.

STORAGE: SYNOLOGY NAS● UNIVERSAL APP SEARCH
02 // THE CONTEXTPROBLEM ➔ FIX
THE CHALLENGE

Manually classifying millions of personal photos stored on a NAS is overwhelmingly time-consuming, making finding specific photos frustrating without manual albums.

THE SOLUTION

Built an automated round-trip workflow: auto-scans the NAS photo library, generates tags with Google Vertex AI, and writes tags back into each photo's EXIF data on the NAS so any photo app can search them.

ZERO LONG PARAGRAPHS
WORKFLOW // NAS → VERTEX AI → EXIF ROUND TRIP
SYNOLOGYSTATUSLANDS923+01020304IPHONEAUTO-SYNC+ DROP PHOTOSBATCH INGESTWEB / MACDESKTOP DAMAUTO-UPLOAD ➔ INGESTFILE SYNC ➔AI TAGGERWATCHER DAEMON• Detects untagged batch● INOTIFY EVENT OK⚡ XAVIER'S SKILLDATA EXTRACTION GUIDE01. Parse Camera EXIF & GPS02. Downsample & Vision Blueprint03. Strict JSON Schema FormattingTEACHES AI HOW TO EXTRACT ➔ENRICHEDGOOGLE VERTEX AIGEMINI 3.8 VISIONMultimodal Visual Scene Analysis🏛️ CUHK🎓 GRADUATION🌿 CAMPUS● 12 SMART TAGS GENERATEDTAGS ➔ EXIF METADATA WRITE-BACK TO NAS ➔ INSTANT SEARCH ON ANY PLATFORM
VERTEX VISION · EXIFTOOL INJECTTAGS IN EXIF · ANY PLATFORM ✓
02 // XAVIER'S ROLE & HOW HE BUILT ITAI SYSTEMS ARCHITECT & DEVELOPER

CASE STUDY ROLE: AI SYSTEMS ARCHITECT & DEVELOPER

Facing the exhausting challenge of manually classifying millions of his own personal photos, Xavier found a way to solve it with automation by engineering a round-trip NAS tagging pipeline.

The background service automatically scans the photo library stored on his Synology NAS. Each photo is forwarded to Google Vertex AI to analyze the visual scene and generate rich, descriptive semantic tags. The generated tags are then written directly back into the photo's own EXIF metadata headers on the NAS using ExifTool.

Because the tags live inside the image files' EXIF itself rather than locked in an external database, any platform, operating system, or photo application can effortlessly search and filter the entire library afterwards. The daemon operates securely in Docker on the NAS, accessible remotely via Cloudflare Zero Trust.

WHAT XAVIER DIRECTLY DELIVERED & SHIPPED
Engineered an automated NAS daemon that auto-scans Xavier's personal photo library without manual intervention
Integrated Google Vertex AI to inspect personal photos and generate rich semantic tags across scenes and subjects
Built the round-trip writeback pipeline using ExifTool to embed tags directly into each photo's EXIF data on the NAS
Ensured tags persist in EXIF so any photo app, operating system, or search tool can effortlessly search and filter the library
Eliminated the tedious pain point of manually classifying millions of personal photos through end-to-end automation
Configured lightweight Docker Compose containerization on Synology NAS with subtle Cloudflare Zero Trust remote access
HANDS-ON OWNERSHIPDIRECT ARCHITECT & BUILDER
03 // LANGUAGES & TOOLS USED
LANGUAGES & CORE RUNTIME
Python 3.9+BashJSON
FRAMEWORKS & ENGINES
FastAPIExifToolDocker Compose
TOOLS, AI MODELS & APIS
Google Vertex AIExifToolSynology Container ManagerCloudflare Zero Trust (subtle access)
STORAGE & INFRASTRUCTURE
Synology NAS Photo StorageEmbedded EXIF MetadataUniversal Photo App Search
CASE STUDY TECH ECOSYSTEM
Python 3.9+Google Vertex AIExifToolEXIF / IPTC MetadataSynology NASDocker ComposeCloudflare Zero Trust
TECH ARCHITECTURE● FULLY FUNCTIONAL
04 // HOW THE SYSTEM WORKS

HOW THE ROUND-TRIP AUTO-TAGGING WORKS

4-STEP PIPELINE
STEP 01
01/04

NAS Auto-Scanning

The automated daemon scans Xavier's personal photo library stored on the Synology NAS.

StatusActive
STEP 02
02/04

Vertex AI Tagging

Photos are passed to Google Vertex AI for intelligent visual recognition and descriptive tag generation.

StatusActive
STEP 03
03/04

EXIF Writeback to NAS

Generated tags are written directly back into each photo's EXIF metadata headers on the NAS.

StatusActive
STEP 04
04/04

Universal Photo Search

Because tags live in EXIF headers themselves, any platform or photo app can instantly search and filter the library.

StatusActive