by Dalpamon Inc.

AI Digital Asset Management · Textile

AI Digital Asset Management for textile manufacturing.

design AI archive

10,000+ samples. Searchable in seconds.

Imagine

design AI archive

What if

A designer could find any fabric sample in 3 seconds?

By texture. By weave. By fiber composition. By season. Without walking to the factory floor.

"Soft viscose-cotton blend, lightweight, suitable for spring jackets"

"Show me last spring's bestselling weaves"

"Texture similar to this swatch but in earth tones"

"What did we ship to client X in 2024 Fall?"

Plain language. Visual previews. No more walking the archive.

The Problem

design AI archive

Today, finding a sample means walking, asking, remembering.

Watch what happens when a designer needs to find samples matching a client's request today.

Scenario"Client wants a swatch like last season's, but in muted tones."

Designer

Senior Designer

Factory Floor

Excel

Receives ask

Recalls from memory

Walks to archive

Old tracking file

More searching

Suggests alt

Emails to client

6 hand-offs · knowledge in one person's head · 2-3 hours per find.

The Solution

design AI archive

A searchable brain above your archive.

Three capabilities turn a physical archive into something a designer can actually query.

Auto-tagging

Every fabric sample is tagged by AI — texture, weave, fiber, color, season — automatically, with no designer keystrokes.

Semantic search

Designers query in natural language or by reference image. The AI returns ranked matches with visual previews — in seconds.

Knowledge persistence

Senior designer's tacit knowledge gets codified into searchable signals. The system survives staff turnover.

Architecture

design AI archive

A layer above your physical archive.

The physical archive stays where it is. The Design AI Brain reads, tags, and indexes. Four modules read from it.

Search

Semantic + visual

· Natural language
· Reference image
· Hybrid ranking

Tag Curator

AI vocabulary tuning

· Custom textile terms
· Senior review queue
· Continuous learning

Designer Workspace

Studio · variations · output

· Collections by season
· AI-generated variations
· BMP + work order export

Analytics

Archive usage insights

· Hot styles by season
· Client preference trends
· Reuse rate per designer

AI Pipeline

Design AI Brain

· Vision tagging· Vector embeddings (pgvector)· Color profile + CIEDE2000

AI-tagged sync

Physical archive · Untouched

Existing samples · shared drives · binders

· Stays the source of truth· Zero schema changes· No migration, no cutover

AI Search

design AI archive

Ask the archive in plain Korean. Get matched designs in seconds.

Designers query naturally — "flower patterns for spring" — and the archive returns ranked matches with thumbnails, tags, and follow-up suggestions. Threaded history per designer.

archive.design-ai / ai-search

LIVE

Natural-language query

Korean or English, conversational.

Ranked visual matches

Thumbnails + tags + similarity score.

Threaded history

Every conversation kept and re-queryable.

Per-designer context

System learns each user’s style over time.

Pattern Editor

design AI archive

From design intent to production-ready pattern.

Designers manipulate the knit grid directly — palette, weave, cuff/heel/toe layers — with live preview at production scale. Output is the file the factory loom needs.

archive.design-ai / pattern-editor

LIVE

Direct grid manipulation

Palette · weave · cuff · heel · toe layers.

Factory-ready output

Files map directly to loom configurations.

2D / 3D Visualizer

Front · back · left · right preview before loom commit.

Sales × Customer

design AI archive

Walk into the customer's office. Pull up every record on the spot.

Sales reps used to scramble through binders and shared drives before customer meetings. With the archive, every past design, color palette, sample file, and production note is one search away — on a tablet, in the meeting room.

archive.design-ai / design-detail

LIVE

Before

Prep hours in binders & drives
"Let me get back to you"
Missed past designs lose deals

After

Tablet · 1 search · all history surfaces
Answers in-meeting, not next day
Reference past wins, close faster

One search, full record

Metadata, tags, palette, files, attachments — all on one screen.

Two Journeys

design AI archive

How designers actually experience it.

Junior Designer

Learns the archive

BEFORE — without archive

First sample request

⚠Shadow senior

⚠Ask senior again

⚠Walk the floor

⚠Wait days for answer

AFTER — with archive

First sample request

Searches archive solo

Sends 5 candidates to client

Time to independenceMonths → Days

Senior Designer

Releases tacit knowledge

BEFORE — without archive

Every request routed in

⚠Recalls from memory

⚠Becomes bottleneck

⚠Knowledge locked in head

⚠Risk of staff turnover

AFTER — with archive

Reviews AI tags weekly

Trains the system once

Knowledge scales beyond them

Capacity per weekBottleneck → Multiplier

AI Capabilities

design AI archive

Production AI capabilities, specifically tuned for textile.

Capability	Technique	Textile-specific tuning
Vision tagging	Claude Vision + custom textile vocabulary	Trained on the customer's internal terms — weave types, fiber grades, finishing techniques.
Semantic similarity	Image + text embeddings (Voyage AI / Bedrock)	1024-dim vectors stored in pgvector. IVFFlat indexing for millisecond similarity search.
Natural language search	Claude + Korean textile glossary	Handles both Korean and English queries. Resolves trade slang to canonical tags.
Color extraction	Color quantization + Pantone matching	CIEDE2000 perceptual color matching. Auto-maps to actual yarn color codes.
Reference-based variation	SDXL + ControlNet	Generates 5 variations per request while preserving the fabric's core structure.

All AI calls routed through AWS Bedrock — Korean data residency by default.

Sample Pipeline

design AI archive

How a sample becomes searchable.

Six steps run automatically the moment a sample is photographed. No designer keystrokes required.

Capture

Photographed + scanned

Extract

Image + color profile

Analyze

Claude Vision: texture, weave, fiber

Enrich

Match the house vocabulary

Embed

Vector embedding

Index

pgvector archive

Input

SW-2026-S-417.jpg

2,481 × 3,508 · 4.2 MB

upload · 2026-05-20 14:32

Auto

≈ 30 sec

per sample

Output · indexed record

SEARCHABLE

textureknit · jacquard

weavefloral motif

fibercotton 60 · viscose 40

palettepastel · spring

categorywomen's quarter

embeddingvec(1536)

→ Now retrievable by natural-language query, similarity search, or any combination of tags.

Existing archive of 5,000–15,000 designs batch-migrated and embedded at project start — same pipeline, run in parallel.

Rollout

design AI archive

Capabilities ship in phases — each one independently usable.

The archive is built so designers gain value at every phase — not just at the end. Foundation is live; AI engine and operations follow.

Live

Design Studio

Designers create directly on the digital archive
Patterns, logos, solid colorways for every sock category
Production-ready output for the loom

Gallery, storage, and unified search across designs
Existing archive migrated in batch
Linked back into Studio for instant reuse

Coming

AI Engine

Semantic search · visual similarity matching
AI variation generation from existing designs
Designer feedback loop trains the system

Final

Workflow & Go-Live

Design request intake from sales and clients
Designer inbox · factory dispatch automation
Production environment fully operational

Outcomes

design AI archive

What changes when knowledge is searchable.

Faster design cycles

Designers find matches in seconds. Prototypes ship in days, not weeks. Client approval rate climbs because options are presented faster.

Metric15-30 min → 3 seconds per find

Knowledge survives turnover

Senior designer's tacit knowledge codified into AI tags and embeddings. When staff move on, the archive doesn't lose anything.

MetricMonths → Days onboarding

Cross-team visibility

Sales, design, production, and clients all reference the same archive. No more "did anyone make something like this before?" emails.

Metric1 source of truth

Data Security & IP

design AI archive

Your samples. Your competitive edge.

Three questions every textile maker asks before letting AI touch their archive.

"Where is the data hosted?"

AWS Seoul region (ap-northeast-2) — Korean data residency by default. Same standards as Korean banking infrastructure. On-prem deployment available on request.

"Who can see the archive?"

Per-user roles. Auditable access logs. Senior staff can lock specific sample collections — for example, exclusive client work hidden from junior designers.

"Could competitors access it?"

Isolated tenant database — row-level security plus separate schemas. The AI engine is shared (the brain). Your data is sealed (the memory). They never cross.

Shared AI engine · Design AI Archive Brain

code · prompts · embedding models · deployed once

queries scoped by tenant

Anchor customer

isolated DB · row-level security

Tenant B (future)

isolated DB · row-level security

Tenant C (future)

isolated DB · row-level security

never cross

Get Started

design AI archive

Built and operated by Dalpamon Inc.

Let's talk knowledge systems.

Every vertical with deep institutional knowledge has the same problem shape. We start with a discovery call to map yours.

simon@dalpamon.com · dalpamon.com

Download one-pager (PDF)

design AI archive

A designer could find any fabric sample in 3 seconds?

Today, finding a sample means walking, asking, remembering.

A searchable brain above your archive.

Auto-tagging

Semantic search

Knowledge persistence

A layer above your physical archive.

Search

Tag Curator

Designer Workspace

Analytics

Ask the archive in plain Korean. Get matched designs in seconds.

From design intent to production-ready pattern.

Walk into the customer's office. Pull up every record on the spot.

How designers actually experience it.

Production AI capabilities, specifically tuned for textile.

How a sample becomes searchable.

Capture

Extract

Analyze

Enrich

Embed

Index

Capabilities ship in phases — each one independently usable.

Design Studio

Asset Layer

AI Engine

Workflow & Go-Live

What changes when knowledge is searchable.

Faster design cycles

Knowledge survives turnover

Cross-team visibility

Your samples. Your competitive edge.

Let's talk knowledge systems.